CIGUATERA IN FLORIDA KEYS PATCH REEFS: BIOGEOGRAPHIC INDICATORS OF GAMBIERDISCUS DENSITY AND TEMPORAL ABUNDANCE (CFP:BIG DATA) A Thesis Presented to The Faculty of the College of Arts and Sciences Florida Gulf Coast University In Partial Fulfillment of the Requirement for the Degree of Master of Science By Meghan Elizabeth Hian 2018
78
Embed
CIGUATERA IN FLORIDA KEYS PATCH REEFS: BIOGEOGRAPHIC ...fgcu.digital.flvc.org/islandora/object/fgcu:31535... · Ciguatera fish poisoning (CFP) is a global public health concern that
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
CIGUATERA IN FLORIDA KEYS PATCH REEFS:
BIOGEOGRAPHIC INDICATORS OF GAMBIERDISCUS
DENSITY AND TEMPORAL ABUNDANCE (CFP:BIG DATA)
A Thesis
Presented to
The Faculty of the College of Arts and Sciences
Florida Gulf Coast University
In Partial Fulfillment
of the Requirement for the Degree of
Master of Science
By
Meghan Elizabeth Hian
2018
APPROVAL SHEET
This thesis is submitted in partial fulfillment of the
requirements for the degree of
Master of Science
Meghan Elizabeth Hian
Approved:
Dr. Michael ParsonsCommittee Chair / Advisor
Dr. Michael Savarese
Dr. S. Gregory Tolley
The final copy of this thesis has been examined by the signatories, and we find that both the content and the form meet acceptable
presentation standards of scholarly work in the above mentioned discipline.
ABSTRACT
Ciguatera fish poisoning (CFP) is a global public health concern that is associated with
Gambierdiscus, a genus of harmful algae found in coral reef environments that includes species
known to produce toxins (ciguatoxins). Outbreaks of CFP have often been linked to elevated
abundance of Gambierdiscus cells and disturbance-related degradation of coral reefs. However,
the influence of human activities on CFP risk, both directly and indirectly within the broader
context of reef health, has yet to be defined for highly exploited patch reefs in the Florida Keys.
The objectives of this study were to define spatial and temporal patterns in reef health and
Gambierdiscus abundance across the three regions (Upper, Middle, Lower), to determine
whether the drivers of those patterns were natural or anthropogenic, and to identify
biogeographic indicators of risk. To address these objectives, this study combined field sampling
with a “big data” approach to spatial analysis. Six patch reefs (two per each of three regions)
were selected as study sites from existing research stations. Datasets from long-term monitoring
of benthic cover, fish species abundance, land use, and water quality were compiled and analyzed
in ArcGIS to characterize the ecological context of each site. Analysis of samples of host
macroalgae collected from all study sites biannually revealed that Gambierdiscus cell densities
were consistently highest in the Upper Keys and lowest in the Middle Keys, regardless of season.
Conversely, reef health was lowest in the Upper Keys and improved along a gradient to the Lower
Keys. Multivariate analysis of site similarity indicated that this regional pattern was driven more
strongly by grazing than substrate availability. Additionally, there is evidence that human
activities have an indirect influence on CFP risk through reef health, as well as through
overfishing, and the destruction of inshore habitats like seagrass and mangroves. Due to a strong
positive correlation with cell densities, this study suggests that mangrove cover could be useful
as a biogeographic indicator of potential CFP risk. Whereas surgeonfish, with a strong negative
correlation with cell densities, could indicate the actual flow of toxins into higher trophic levels.
The concordance of high regional risk and high population density necessitates continued
monitoring of fish in those areas and the development of more comprehensive predictor of
potential CFP outbreaks.
Acknowledgements
This study was funded in part by NOAA CiguaHAB Award # NA11NOS4780028.
I would like to express my sincere gratitude to my advisor Dr. Parsons for his continuous
support of my work and research. It was always a dream of mine to study dinos, and that dream
came true when I joined the Parsons benthic dinoflagellate research lab. The dinos that I got to
study were just a bit smaller by an order of magnitude or so. Regardless, I was extremely
fortunate to work with such a respected researcher who also turned out to be a cool boss, to
sample down in the keys and, even when faced with immediate tire blow-outs, extreme sun,
equipment loss, and one small fire, to still do science! I would like to thank my committee
members Dr. Tolley and Dr. Savarese for their time and support, both academically and
professionally. I feel incredibly lucky to be part of CWI and have thoroughly enjoyed working
with the faculty, staff and students as both a student and a colleague.
Thanks to the lab, especially Adam Catasus, Jeff Zingre, Nick Culligan, Jesse Elmore, Alex Leynse,
Anne Smiley, Andrea James, and Katie Ribble for making our eventful trips so enjoyable, and for
all of their help with collecting data, processing samples, and creative repairs. Thanks to my
fellow counters, Sammi Blonder and Jessica Schroeder, for making good musical choices in the
microscope room. And thanks to Dr. Venture for betting on me.
I cannot thank my husband Ryan enough for the endless encouragement, patience, and support
through the entire endeavor. Thanks to Sherman, Stanley, and Schnitzel for all of their “help”
with studying and writing, to my nephews Elijah and Elliot for adding the Moana soundtrack, to
my sisters Jenn and Allie for believing in me (and putting up with my weirdness), and to my
father for inspiring me to never give up.
Table of Contents 1 INTRODUCTION ................................................................................................................................................ 3
1.1 History of CFP ........................................................................................................................................................... 4
1.2 Biogeography of Gambierdiscus ............................................................................................................................... 5
1.3 Dynamics of a Harmful Algal Bloom ......................................................................................................................... 7
1.4 Dynamics of the Reef Environment ........................................................................................................................ 11
1.5 Anthropogenic Factors in the Florida Keys ............................................................................................................. 13
1.6 Link to CFP risk—Trophic Transfer .......................................................................................................................... 15
1.7 Research Objectives................................................................................................................................................ 16
2.1 Description of Study Sites ....................................................................................................................................... 19
2.4 Measures of Reef Health ........................................................................................................................................ 23
2.6 Geodatabase and GIS Synthesis .............................................................................................................................. 25
2.7 Data Analysis .......................................................................................................................................................... 26
2.8 Relation to Gambierdiscus Density & Temporal Abundance .................................................................................. 29
3.1 Patterns in Land Use ............................................................................................................................................... 32
3.2 Patterns in Water Quality ....................................................................................................................................... 35
3.3 Patterns in Key Fish Species Assemblages .............................................................................................................. 37
3.4 Patterns in Benthic Cover ....................................................................................................................................... 39
3.5 Patterns in Reef Health ........................................................................................................................................... 42
3.6 Patterns in Gambierdiscus Cell Densities ................................................................................................................ 44
3.7 Biotic and Environmental Correlations ................................................................................................................... 46
4.1 Population Dynamics of Gambierdiscus spp. .......................................................................................................... 49
4.2 Land Use as a driver ................................................................................................................................................ 52
4.3 Anthropogenic Influence on the reef ..................................................................................................................... 54
TABLE 2.7.1 SIRHI THRESHOLD VALUES FROM MCFIELD ET AL. (2011) ..................................................................................... 27
TABLE 2.7.2 FISH BIOMASS CONVERSIONS ............................................................................................................................ 28
TABLE 3.1.1 DISCRIMINATING LAND USES BY AREA WITHIN 10 KM (SIGNIFICANT CONTRIBUTIONS IN BOLD) ........................................ 33
TABLE 3.1.2 DISCRIMINATING LAND USES BY PERCENT COVER WITHIN 10 KM OF STUDY SITES (SIGNIFICANT CONTRIBUTIONS IN BOLD) .... 35
TABLE 3.2.1 TEMPERATURE AND SALINITY AT BOTTOM (B) OF STUDY SITES FROM 2010-2015 ....................................................... 36
TABLE 3.2.2 AVERAGE WATER QUALITY WITHIN 5KM AND ANTHROPOGENIC FACTORS WITHIN 10KM OF STUDY SITES. ........................ 37
TABLE 3.3.1 DISCRIMINATING FISH SPECIES BY ABUNDANCE (SIGNIFICANT CONTRIBUTIONS IN BOLD) ................................................ 38
TABLE 3.4.1 AVERAGE ABUNDANCE OF DISCRIMINATING BENTHIC COVER GROUPS (SIGNIFICANT CONTRIBUTIONS IN BOLD) .................. 41
TABLE 3.5.1 BENTHIC COVER AND KEY FISH BIOMASS VALUES AND HEALTH SCORES BY STUDY SITE AND REGION ................................ 43
TABLE 3.7.1 BEST RESULTS FOR MULTI-CORRELATIONS WITH 5 OR FEWER VARIABLES ................................................................... 46
TABLE 3.8.1 REGIONAL ASSESSMENT OF POTENTIAL TOXIN AVAILABLE TO THE CONSUMER PER G COMMERCIAL FISH ............................. 47
FIGURES
FIGURE 2.2.1 MAP OF STUDY SITES IN EACH REGION; TP=TWO PATCHES, BF=BURR FISH, RR=RAWA REEF, DR=DUSTAN ROCKS, WL=
WONDERLAND, WW= WEST WASHERWOMEN ............................................................................................................. 21
FIGURE 3.1.1 CLUSTER ANALYSIS OF SQUARE ROOT TRANSFORMED AREAS OF LAND USE WITHIN 10 KM OF SITES. SAMPLES CONNECTED BY
RED LINES ARE NOT SIGNIFICANTLY DIFFERENTIATED BY SIMPROF. REGION 1=UPPER KEYS, 2=MIDDLE KEYS, 3=LOWER KEYS. ... 32
FIGURE 3.1.2 CLUSTER ANALYSIS OF SQUARE ROOT TRANSFORMED PERCENT COVER OF LAND USE WITHIN 10KM OF SITES. SAMPLES
CONNECTED BY RED LINES ARE NOT SIGNIFICANTLY DIFFERENTIATED BY SIMPROF. REGION 1=UPPER KEYS, 2=MIDDLE KEYS, 3=LOWER KEYS. ...................................................................................................................................................... 34
FIGURE 3.3.1 CLUSTER ANALYSIS OF SQUARE ROOT TRANSFORMED DENSITIES OF KEY FISH SPECIES WITHIN 5KM OF SITES. SAMPLES
CONNECTED BY RED LINES ARE NOT SIGNIFICANTLY DIFFERENTIATED BY SIMPROF. REGION 1=UPPER KEYS, 2=MIDDLE KEYS, 3=LOWER KEYS. ...................................................................................................................................................... 37
FIGURE 3.4.1 CLUSTER ANALYSIS OF BRAY-CURTIS SIMILARITY ON SQUARE ROOT TRANSFORMED BENTHIC COVER DATA. SAMPLES
CONNECTED BY RED LINES ARE NOT SIGNIFICANTLY DIFFERENTIATED BY SIMPROF. REGION 1=UPPER KEYS, 2=MIDDLE KEYS, 3=LOWER KEYS. ...................................................................................................................................................... 39
FIGURE 3.4.2 CLUSTER OVERLAY ON NMDS; REGION 1=UPPER KEYS, REGION 2=MIDDLE KEYS, REGION 3=LOWER KEYS................ 40
FIGURE 3.5.1 BENTHIC COVER ACROSS ALL STUDY SITES FROM 2010 TO 2015; UPPER KEYS=TWO PATCHES, BURR FISH; MIDDLE
FIGURE 3.6.1 SEASONAL CELL DENSITY OF GAMBIERDISCUS SPP. ON H. GRACILIS WITH STANDARD ERROR; CROSS-HATCHED BAR REPRESENTS
MISSING DATA ......................................................................................................................................................... 44
FIGURE 3.6.2 MEAN CELL DENSITY OF GAMBIERDISCUS SPP. ON H. GRACILIS WITH STANDARD ERROR ............................................... 45
3
1 INTRODUCTION
Ciguatera Fish Poisoning (CFP), once thought to be a tropical disease confined within the latitudes
of 35°N and 35°S, is emerging as a global public health concern that may affect as many as half a
million people each year (Bravo et al., 2015; Radke et al., 2015; Roeder et al., 2010; Lehane &
Lewis, 2000). CFP is an illness caused by ingesting fish—typically reef fish—that have accumulated
naturally-occurring toxins (ciguatoxins). Unlike other forms of food-borne illness, CFP is not
caused by improper storage, handling, or preparation of the fish (Thompson et al., 2017).
Ciguatoxins (CTXs) are lipid-soluble and heat stable, meaning that a fish may remain unsafe to
eat after extended periods of freezing as well as after cooking (Friedman et al., 2017).
Additionally, CTXs are colorless, tasteless, and odorless, making them difficult to detect in a raw
fillet or a prepared meal. Once ingested, the specific set of symptoms experienced due to CTX
exposure appear to depend upon the location from which the fish was obtained (i.e., Pacific
Ocean versus Caribbean Sea versus Indian Ocean). This variability is likely attributable to the
chemical structure of CTX, which differs slightly by geographic region (Friedman et al., 2017;
Lehane & Lewis, 2000). For example, intoxication by C-CTX 1 (found in the Caribbean) is
characterized by acute gastrointestinal symptoms, such as diarrhea, nausea, and vomiting, that
occur within several hours of eating ciguatoxic fish, which are later followed by neurologic
symptoms, such as tingling sensations, itchy skin, and cold allodynia (reversal of hot-cold
sensation), and occasionally cardiac symptoms, such as hypotension (low blood pressure) and
reduced heart rate (FDA, 2011; Friedman et al., 2017). Generally, acute symptoms resolve within
several days but may be followed by chronic fatigue and recurrence of neurologic symptoms
4
(Friedman et al., 2017). The U.S. Food and Drug Administration (FDA) has established the action
level for Caribbean toxin levels at 0.1 ppb for C-CTX-1; however, a rapid test has yet to be
validated to screen fish at this low concentration (2011). Therefore, advisories for CFP and control
of potentially harmful seafood is still largely reactionary to reports of illness and anecdotal
evidence of “hotspots” to avoid.
1.1 History of CFP
The term ciguatera first appeared in a book published in Havana, Cuba in 1787 and was initially
used to describe an illness contracted after eating a certain type of sea snail (Turbo pica),
(Scheuer, 1994). Later, the definition was refined to refer specifically to an intoxication caused
by the ingestion of coral reef fishes. Although accounts of CFP date back to the 1500s in the
Americas, and records from Captain James Cook in 1774 describe a probable case in the Pacific
(Scheuer, 1994), there is evidence to suggest that it affected coastal societies much earlier.
According to Rongo et al. (2009), CFP may have even induced the great oceanic voyages of the
Polynesians from A.D. 1000 to 1450. Research on CFP in the U.S. was pioneered by a Navy
physician named Bruce Halstead who took an interest in the phenomenon during WWII while in
the Pacific theatre (Scheuer, 1994). At this point, the exact source of the toxin was still unknown,
though there was mounting evidence of a trophic connection with benthic algae (Randall, 1958;
Halstead, 1965). After several decades of investigation, microscopic algae from the genus
Gambierdiscus (Adachi and Fukuyo, 1979) were definitively linked to CFP by Yasumoto et al.
(1977; originally called Diplopsalis) and confirmed to produce the ciguatera toxins (Bagnis et al.,
1980; Lehane & Lewis, 2000). Once thought to be monospecific (G. toxicus), the genus
5
Gambierdiscus now comprises over a dozen described species of morphologically similar armored
photosynthetic dinoflagellates (Fraga & Rodriguez, 2014; Richlen et al., 2008). Dinoflagellates are
a diverse group of phytoplankton that is distinguished by their two unequal flagella, which aid in
the motility of the cells. Perhaps due to their ability to swim, dinoflagellates are able to flourish
under a diverse set of environmental conditions and have an extensive fossil record dating back
several hundred million years (Hackett et al., 2004). Unlike their naked counterparts like Karenia
brevis (responsible for red tides in Southwest Florida), armored dinoflagellate cells are covered
by a theca, made up of plates of cellulose or other polysaccharides, that are arranged in distinct
patterns within their membranes (Hackett et al., 2004). Despite the utility of this pattern as a
taxonomic identifier, the small scale of the differences within the genus Gambierdiscus makes it
difficult to identify species using light microscopy alone (Litaker et al., 2010).
1.2 Biogeography of Gambierdiscus
Of the currently described species, two (G. caribaeus and G. carpenteri) have a cosmopolitan
distribution (Litaker et al., 2010), with populations of G. caribaeus representing the most
commonly found species in both the Atlantic and Pacific (Litaker et al., 2009). Likely, the
distribution of these two species is a reflection of their broad temperature tolerance relative to
others in the genus (Tester et al., 2010). Species endemic to the Atlantic include G. belizeanus, G.
carolinianus, G. ruetzleri (now genus Fukuyoa), G. ribotype 2, G. silvae (previously G. ribotype 1),
and G. excentricus (Litaker et al., 2017; Fraga & Rodriguez, 2014; Tester et al., 2013; Litaker et al.,
2010). Aside from G. caribaeus, G. carpenteri and F. yasumotoi, populations of Gambierdiscus in
the Atlantic are phylogenetically distinct from those in the Pacific (G. australes, G. pacificus, G.
6
polynesiensis, and G. toxicus) (Litaker et al., 2010). It has been suggested that this geographic
divergence may be traced back to the period between the Miocene and the Pleistocene when
the closing of the Tethys Sea and the formation of the Isthmus of Panama disrupted the
circumtropical flow of the sea (Rodriguez et al., 2017). Vicariance during this period seems to be
supported by the high biodiversity of Gambierdiscus now found in the Atlantic, Caribbean, and
Gulf of Mexico, often with five or more different species present in the same location (Rodriguez
et al., 2017; Tester et al., 2013). Due to the lack of appropriate genetic markers, little is known
about the population structure of Gambierdiscus species, and studies of connectivity and
dispersal have only recently become possible with newly developed microsatellite methodology
(Sassenhagen & Erdner, 2017; Kuno et al., 2010). Nevertheless, the biogeographic range of
Gambierdiscus appears to be expanding into areas without a history of CFP, such as the northern
Gulf of Mexico (Tester et al., 2013; Villareal et al., 2007), East Asia (Kuno et al., 2010), and the
Canary Islands (Rodriguez et al., 2017; Bravo et al., 2015; Fraga & Rodriguez, 2014), though it is
possible that these areas harbored existing populations that had been previously overlooked or
understudied. Gambierdiscus spp. have been observed rafting on drift algae (Bomber et al.,
1988), making dispersal to new areas possible, especially in areas where artificial reefs can act as
“stepping stones” (Villareal et al., 2007). With dispersal, climate change could also play a role in
the alteration of the biogeographic range, as novel habitats that begin to fall within the
temperature tolerance of particular species could be colonized.
7
1.3 Dynamics of a Harmful Algal Bloom
Generally, Gambierdiscus spp. are found in shallow (<50 m) tropical and subtropical marine reef
habitats characterized by less than 10% of incident light (Litaker et al., 2010), stable salinities
between 29 and 34 ppt (Kibler et al., 2012; Parsons et al., 2010), annual water temperatures
between 18° and 33°C (Litaker et al., 2010; Tester et al., 2010; Chateau-Degat et al., 2005; Chinain
et al., 1999; Hales et al., 1999), and abundant natural or artificial reef substrates (Parsons et al.,
2017; Villareal et al., 2007). As motile cells, they have been observed within the water column
and swimming in the epibenthos (Parsons et al., 2011; Nakahara et al., 1996); however,
Gambierdiscus cells are predominately epiphytic and often attach to organic substrates such as
macrophytes and algal turfs (Parsons et al., 2011). Although attachment to inorganic structures
has also been noted (Parsons et al., 2017; Villareal et al., 2007), most studies have focused on
macroalgal substrates, as those are the most likely vector of CTX (via herbivory) into the food
web (Rains & Parsons, 2015). Additionally, previous studies have suggested that macroalgal hosts
exude substances that can either stimulate or inhibit Gambierdiscus growth (Parsons et al., 2011),
though the role of exudates in attachment behavior and substrate may be masked by the effects
of other environmental variables. For example, attachment behavior may also be affected by
changes in light conditions, local physical disturbance, or the presence of bacteria or other
epiphytes that drive competition and may play a role in growth inhibition (Sakami et al., 1999;
Nakahara et al., 1996; Tosteson et al., 1989). Further, differences in epiphytic behavior in relation
to a variety of macroalgal substrates could be attributed to interspecific host preferences
exhibited by the Gambierdiscus themselves (Rains & Parsons, 2015). Likely, this variability in
8
behavior and preference is a function of differential environmental tolerances among species in
the genus.
Gambierdiscus cells grow at a relatively slow rate of about one division every three days
(Lehane & Lewis, 2000). Although growth has been shown to be influenced by temperature,
salinity, and irradiance in the lab (Xu et al., 2016), the slow growth rate introduces complexity to
the relationship with bloom dynamics in the field due to the temporal scale of local
environmental variability. Typically, researchers have used cell densities, or an elevated
abundance of Gambierdiscus cells present on their macroalgal hosts, as a proxy for CFP risk
(Parsons et al., 2010). Although Gambierdiscus spp. are considered to be a type of harmful algal
bloom (HAB) (Grattan et al., 2016; Anderson et al., 2008), it has been difficult to characterize the
threshold at which a bloom occurs. The literature has suggested that a bloom occurs when the
local density exceeds 1,000 cells g-1 wet weight algae (Litaker et al., 2010). However, abundance
alone may not truly be indicative of the potential for a CFP outbreak, as there is variability in CTX
production within the genus Gambierdiscus. Although CTX production has been shown to vary
with environmental factors, such as temperature, salinity, light, and nutrients (Chinain et al.,
2010; Morton et al., 1992; Holmes et al., 1991; Bomber et al., 1988), no consistent pattern of
seasonality was observed across regions, and no correlation was found between toxicity of these
blooms and their biomass (Chinain et al., 1999). Because many of these studies were done when
Gambierdiscus was still considered to be a single species, the observed variability in toxicity in
relation to environmental factors may be confounded with other interspecific differences in
biology, physiology, and ecology (Parsons et al., 2012). Therefore, the risk of toxicity in a given
area is understood to depend more on the clonal nature of cells within the local populations than
9
seasonal or environmental factors (Chinain et al., 1999), due to the fact that the ability to produce
CTXs appears to be genetically determined (Chinain et al., 2010; Roeder et al., 2010; Richlen et
al., 2008).
Of the seven species identified in the Atlantic, all have been reported to produce some
amount of CTXs, with the slowest growing species observed to exhibit the highest toxicity per cell
(Litaker et al., 2017; Tester et al., 2013; Chinain et al., 2010). A recent study characterized
Gambierdiscus excentricus as highly toxic, Gambierdiscus silvae and Gambierdiscus ribotype 2 as
and total phosphorus (≤0.2 micromolar). Using IBM SPSS Statistics (v23), the mean, maximum,
and minimum descriptive statistics were calculated by parameter for each site. Water quality was
compared by site and region with independent sample Mann-Whitney tests and against EPA
strategic targets with one-sample T-tests. Finally, the relationship between patterns of
Gambierdiscus abundance and site conditions were explored in SPSS through correlations with
environmental variables identified by the aforementioned SIMPER and BEST analyses. The
significance level α was set at 0.05 for all statistical tests.
2.9 CFP Risk Calculation
Under the assumption that macroalgal densities correlate across species (Parsons et al., 2017),
cell densities (no. g-1 ww macroalgae) were extrapolated to number of cells per reef (300 m2).
These figures were based on percentages of benthic cover and conversions of macroalgal cover
to biomass using linear regressions developed by Parsons et al. (2017). From reef-scale cell
31
enumeration, an estimate of total micrograms of toxin per reef was also calculated using a
generic coefficient to represent toxin content per cell. Although the precise mechanisms of
trophic transfer require further study, this model operated under the assumption that all toxin
was evenly consumed by the herbivorous fish. Finally, a ratio of trophic transfer (herbivorous
biomass/commercial biomass) was established to characterize the potential amount that could
reach the consumer. Although the pathways are likely more complex, the generic coefficients
used in concert with the data in this study is sufficient to assess and compare the risk among
regions.
32
3 RESULTS
3.1 Patterns in Land Use
SIMPER analysis indicated that the average similarity between sites was 98.09% in the Upper
Keys, 86.87% in the Middle Keys and 80.33% in the Lower Keys, with significant structural
differences apparent between regions. CLUSTER analysis of the square root transformed areas
from the Monroe County Land Use and Cover dataset revealed a pattern of transition from Lower
to Upper Keys (Fig 3.1.1).
Figure 3.1.1 CLUSTER analysis of square root transformed areas of land use within 10 km of sites. Samples connected by red lines are not significantly differentiated by SIMPROF. Region 1=Upper Keys, 2=Middle Keys, 3=Lower Keys.
33
Average dissimilarity in land use area between regions was highest between the Lower and
Middle Keys at 27.28%, lower still between the Lower and Upper Keys at 24.00%, and lowest
between the Middle and Upper Keys at only 12.31%. Average abundances of Scrub Mangroves
and Developed Land within 10 km (Table 3.1.1) were defined as discriminating factors for the
Lower Keys relative to the other regions. Additionally, Salt Marsh was identified area as a key
contributor to the dissimilarity between Upper and Lower Keys, with greater cover associated
with the Lower Keys.
Table 3.1.1 Discriminating land uses by area within 10 km (significant contributions in bold)
CLUSTER analysis was also performed on square-root-transformed percent cover of each land
use category. Between-region differences were greater than within-region differences, as each
pair of sites showed no significant structural differentiation (Fig. 3.1.2). SIMPER analysis
reinforced that the Upper Keys again had the highest resemblance and an average similarity
between sites of 98.63%, while the Middle Keys with an average similarity of 92.74% and the
Lower Keys with an average similarity of 87.44% followed a trend of decreasing resemblance.
34
Figure 3.1.2 CLUSTER analysis of square root transformed percent cover of land use within 10km of sites. Samples connected by red lines are not significantly differentiated by SIMPROF. Region 1=Upper Keys, 2=Middle Keys, 3=Lower Keys.
Average dissimilarity in the percent cover of different land uses was highest at 28.99% between
the Lower and Middle Keys, and was similarly high at 27.28% between the Lower and Upper Keys.
Setting the Lower Keys apart, an average dissimilarity of only 11.18% was seen between the
Middle and Upper Keys. Low salt marsh density in the Upper Keys contributed to 19.30% of the
dissimilarity with the Middle Keys and to 14.83% of the dissimilarity with the Lower Keys (Table
3.1.2). Low cover of exotics and high hammock cover in the Upper Keys also contributed to its
differentiation from the Middle Keys. Scrub Mangrove and Developed Land again distinguished
the Lower Keys from the rest of the study area. The high scrub mangrove cover in the Lower Keys
accounted for 26.79% of the dissimilarity with Upper Keys and 28.26% of the dissimilarity with
the Middle Keys. Furthermore, the lower percentage of developed land near the study sites in
the Lower Keys contributed to between-region dissimilarities.
35
Table 3.1.2 Discriminating land uses by percent cover within 10 km of study sites (significant contributions in bold)
Figure 3.3.1 CLUSTER analysis of square root transformed densities of key fish species within 5km of sites. Samples connected by red lines are not significantly differentiated by SIMPROF. Region 1=Upper Keys, 2=Middle Keys, 3=Lower Keys.
Cluster analysis of the square-root-transformed estimated mean population calculated from the
REEF order of magnitude surveys 20102017 illustrated that key fish species assemblages also
38
follow a generally regional pattern (Fig. 3.3.1). Assemblages from the pairs of sites in the Upper
and Lower Keys do not show any significant structural within-region differences, though their
particular degree of similarity varied between the two regions. There were also significant
differences in the Lower Keys fish populations in relation to those of the Middle and Upper Keys.
SIMPER analysis suggested that key fish species assemblages at the sites in the Upper Keys are
nearly identical with an average between-site similarity of 98.23%. The Lower Keys are shown to
have more within-region variability in their assemblages with an average between-site similarity
of only 74.14%. Within-region variability was even greater in the Middle Keys, with two distinct
structural groupings in the cluster and the lowest average between-site similarity of 69.23%.
Between-region differences in key fish assemblages seemed to follow a gradient from West to
East, as an average dissimilarity of 41.63% characterized the Lower and Middle Keys, an average
dissimilarity of 34.94% was estimated between the Lower and Upper Keys, and an average
dissimilarity of 28.44% defined the Middle and Upper Keys pairing.
Table 3.3.1 Discriminating fish species by abundance (significant contributions in bold)
The differential abundance (Table 3.3.1) of the Striped Parrotfish (Scarus iseri) had the highest
contribution (21.43%) to the dissimilarity between the Lower and Upper Keys. Along with the
Schoolmaster (Lutjanus apodus), the Striped Parrotfish also contributed highly to the dissimilarity
between the Lower and Middle Keys. The dissimilarity between the Middle and Upper Keys was
also marked by the contrasting abundance of the Schoolmaster (L. apodus). The high abundance
of the Mahogany Snapper (Lutjanus mahogoni) and the low abundance Yellowtail Snapper
(Ocyurus chrysurus) in the Upper Keys further defined the regional dissimilarity.
3.4 Patterns in Benthic Cover
Figure 3.4.1 CLUSTER analysis of Bray-Curtis similarity on square root transformed benthic cover data. Samples connected by red lines are not significantly differentiated by SIMPROF. Region 1=Upper Keys, 2=Middle Keys, 3=Lower Keys.
CLUSTER analysis in PRIMER7 of square-root-transformed averages of the CREMP coral point
counts indicated a high level of similarity among benthic assemblages at the regional level (Fig.
40
3.4.1). Aside from one outlier at Burr Fish reef in 2010 and one at Rawa Reef in 2014, benthic
cover was also highly similar over time, with very few significant differences within sites and
within regions.
Figure 3.4.2 CLUSTER overlay on nMDS; Region 1=Upper Keys, Region 2=Middle Keys, Region 3=Lower Keys
A non-metric MDS with an overlay of the previous CLUSTER (Fig. 3.4.2) and SIMPER analyses
confirmed this pattern in benthic cover with an acceptable 2-dimensional stress of 0.12. Based
on the nMDS, average benthic cover was at least 75% similar across all sites and all years between
2010 and 2015. Within regions, a SIMPER analysis revealed that the benthic cover in the Lower
Keys had an average similarity of 86.59% among its sites, which was the highest similarity
observed. Similarity in benthic cover among sites in the Middle and Upper Keys was comparable,
though slightly less than that of the Lower Keys, with average similarities of 84.97% and 84.51%,
41
respectively. Between regions, the highest average dissimilarity (30.99%) was found between the
Upper Keys and Lower Keys, and lowest (19.65%) was found between the Upper and Middle Keys.
Table 3.4.1 Average abundance of discriminating benthic cover groups (significant contributions in bold)
The average dissimilarity between the Middle and Lower Keys (24.38%) was nearly half that of
the other two pairings. SIMPER analysis also distinguished the groups of species that contributed
most to the average dissimilarity between regions. These discriminating species groups were
defined as those that contributed >10% to the average dissimilarity and had a Diss/SD ratio >2.0
(Wakefield et al., 2013).
Analysis of between-region dissimilarity (Table 3.4.1) suggested that the percentages of Stony
Coral and substrate cover contributed most to the differentiation of the Middle Keys from the
Upper Keys and Lower Keys from both other regions. Macroalgae cover also contributed heavily
(>25%) to the distinction of the Lower Keys from the Upper Keys. In addition to the
aforementioned discriminating species groups, the percent cover of Zoanthids uniquely
distinguished the Middle Keys from other regions.
42
3.5 Patterns in Reef Health
Figure 3.5.1 Benthic Cover across all study sites from 2010 to 2015; Upper Keys=Two Patches, Burr Fish; Middle Keys=Rawa Reef, Dustan Rocks; Lower Keys= Wonderland, West Washerwomen.
Stony coral cover followed a gradient from “good” in the Lower Keys, with decreasing cover
moving East-Northeast, to “critical” in the Upper Keys (Table 3.5.1). Inter-annual variability was
relatively low across all sites, though a steeper decline in cover was apparent in the Lower Keys
over the most recent years in the dataset (Fig. 3.5.1). A one-way ANOVA on the log-transformed
percentage of stony coral cover revealed significant differences between each of the three
regions (p < 0.001), as well as between sites in the Lower Keys (p = 0.003). Macroalgal cover
followed a similar trend and was rated as “good” in the Lower Keys; however, the rating jumped
to “poor” in the Middle and peaked at “critical” in the Upper Keys (Table 3.5.1). A relatively high
level of within-region concordance in macroalgal cover was evident in the Middle and Lower
Keys, with little inter-annual variability in the Lower Keys (Fig. 3.5.1). Although macroalgal cover
in the Upper Keys appeared to be more volatile, a one-way ANOVA on the square-root-
0%
15%
30%
45%
2010 2011 2012 2013 2014 2015
Year
Stony Coral CoverTwo Patches
Burr Fish
Rawa Reef
Dustan Rocks
Wonderland
WestWasherwomen
0%
15%
30%
45%
2010 2011 2012 2013 2014 2015
Year
Macroalgal Cover
43
transformed percentage of cover revealed that any within-region differences in cover were not
statistically significant. ANOVA also confirmed that differences in macroalgal cover between the
Lower Keys and both the Middle and Upper Keys were statistically significant (cover was lower at
the Lower Keys sites; p < 0.001), as were those between the Middle and Upper Keys (cover was
higher at the Upper Keys sites; p = 0.027).
Table 3.5.1 Benthic Cover and Key Fish Biomass values and Health Scores by study site and region
Region
Site
Stony Coral
Macroalgae
Herbivorous Fish Biomass
Commercial Fish Biomass
Assessed
Score Cover Score Cover Score g 100m-2 Score g 100m-2 Score
Upper Keys
Two Patches 4.9% 1 16.9% 2 86.33 1 324.22 1 1
Burr Fish 7.9% 2 26.9% 1 81.61 1 312.61 1 1
Middle Keys
Rawa Reef 11.9% 3 12.7% 2 233.72 1 37.53 1 2
Dustan Rocks 13.0% 3 12.4% 2 224.73 1 87.30 1 2
Lower Keys
Wonderland 37.9% 4 1.7% 4 95.52 1 440.50 2 3
W. Washerwomen 26.8% 4 2.2% 4 215.12 1 386.27 1 3
The densities of key fishes were consistently low throughout the entire Florida Keys Reef
Tract (Table 3.5.1), with key herbivorous fish biomass rated at “critical” across all sites. The
biomass of key commercial species was also rated at “critical” at all sites except for Wonderland
in the Lower Keys. Independent samples Kruskal-Wallis tests across regions confirmed that the
differences in the distributions of commercial and herbivorous fish biomass across regions were
statistically significant (p = 0.005; p = 0.027), with the Middle Keys sites containing the highest
and lowest densities of herbivorous and commercial fish species, respectively. Overall, the
Assessed Health Score was consistent within regions and ranged between regions from “critical”
(1) in the Upper Keys to “fair” (3) in the Lower Keys.
44
3.6 Patterns in Gambierdiscus Cell Densities
The temporal abundance of Gambierdiscus cells did not indicate a clear pattern of seasonality at
the study sites (Fig. 3.6.1). In the Upper and Middle Keys, samples collected in winter (February
2017) hosted cell densities that were similar to or slightly higher than those recorded from the
previous summer (August 2016). Cell densities observed in samples from the following summer
(August 2017) followed a similar pattern, but generally had higher densities than those from the
previous two seasons. In the Lower Keys, Halimeda spp. samples could not be collected from
either site during the winter due to the near absence of macrophytes at both sites. This apparent
seasonal low in macroalgal abundance was confirmed by another attempt to sample during
winter at both Lower Keys sites in December 2018. Regardless, statistically significant (p = 0.002)
interannual variability was apparent across all regions through an increase in cell densities
between August 2016 and 2017.
Figure 3.6.1 Seasonal cell density of Gambierdiscus spp. on H. gracilis with standard error; Cross-hatched bar represents missing data
0
10
20
30
40
50
60
70
80
Upper Keys Middle Keys Lower Keys
Cel
ls g
-1 w
w
Seasonality
Aug-16
Feb-17
Aug-17
45
Across all study sites, the highest average cell density of Gambierdiscus spp. was recorded at Burr
Fish reef in the Upper Keys and the lowest average at Rawa Reef in the Middle Keys (Fig. 3.6.2).
Linear mixed model analysis in IBM SPSS Statistics 23 revealed that the square root transformed
mean cell density at Burr Fish reef was significantly higher than that at Two Patches (p = 0.01) in
the Upper Keys, both Rawa Reef (p < 0.001) and Dustan Rocks (p < 0.001) in the Middle Keys, and
West Washerwomen (p < 0.001) in the Lower Keys. The analysis also indicated the significance of
the lower mean cell density at Rawa Reef in the Middle Keys as compared to both Burr Fish (p <
0.001) and Two Patches (p = 0.005) in the Upper Keys, and Wonderland (p = 0.008) in the Lower
Keys. Statistically significant within-region differences in mean cell density were only found in the
Upper Keys.
Figure 3.6.2 Mean cell density of Gambierdiscus spp. on H. gracilis with standard error
On a regional scale, cell densities observed in the Upper Keys were significantly higher in relation
to both the Middle Keys (p < 0.001) and the Lower Keys (p = 0.02). Differences in cell densities
B C ABA BC BC0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
Upper Keys Middle Keys Lower Keys
SQR
T C
ells
g-1
ww
Site Average Cell Density
46
between the Lower Keys and the Middle Keys were less pronounced. Although the Lower Keys
generally had higher cell densities than the Middle Keys, the difference was marginal (p = 0.05).
3.7 Biotic and Environmental Correlations
The BEST analysis in PRIMER7 was utilized to describe how well patterns in land use area and
density described patterns in benthic cover and fish species assemblages (Table 3.7.1).
Table 3.7.1 BEST results for multi-correlations with 5 or fewer variables