Florida State University Libraries

2015 Ecological Effects of Red Grouper (Epinephelus Morio) in Florida Bay Robert Dodge Ellis

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] FLORIDA STATE UNIVERSITY

COLLEGE OF ARTS AND SCIENCES

ECOLOGICAL EFFECTS OF RED GROUPER (EPINEPHELUS MORIO)

IN FLORIDA BAY

By

ROBERT DODGE ELLIS

A Dissertation submitted to the Department of Biological Science in partial fulfillment of the requirements for the degree of Doctor of Philosophy

2015 Robert D. Ellis defended this dissertation on October 7, 2015. The members of the supervisory committee were:

Felicia C. Coleman Professor Directing Dissertation

Markus Huettel University Representative

Emily H. DuVal Committee Member

Brian D. Inouye Committee Member

Thomas E. Miller Committee Member

The Graduate School has verified and approved the above-named committee members, and certifies that the dissertation has been approved in accordance with university requirements.

ii

For Caitlin and Will.

iii ACKNOWLEDGMENTS

This dissertation would not have been possible without the advice, support (financial,

academic, moral, and otherwise), and encouragement of my major advisor Felicia Coleman.

Major thanks must also be paid to my (unofficial) co-advisor Tom Miller, whose willingness to

adopt a fish guy into a terrestrial lab and weekly chats helped smooth the various bumps along

the road and challenged me to become a better ecologist. Thanks to the members of my doctoral

committee: Emily DuVal, Brian Inouye, and Markus Huettel; their contributions to this work and advice throughout helped me think more broadly about what I was doing. I received incredible

guidance and support from Chris Koenig who advised me on nearly every part of this

dissertation and, even more importantly, about how to be a good scientist and colleague.

Professor Bill Herrnkind provided valuable input about lobster biology and about how to be a

good mentor. Mark Butler and Mark Hixon provided encouragement and valuable advice about

lobsters and lionfish, respectively.

My dissertation research was supported by funds provided by the Florida State University

Department of Biological Science, the FSU Coastal and Marine Laboratory, the Sigma Xi

Scientific Research Society, the Guy Harvey Ocean Foundation, the PADI Foundation, and from

the 37 people who supported by project through SciFlies.

The field work in Florida Bay could not have happened without the assistance of an

amazing group of friends and colleagues, including the four undergrads that I mentored through

the FSU Marine Certificate Program – Ashley Carreriro, Meaghan Faletti, Amie Lentner, and

Nate Levine – and all the short timers, dive buddies, and boat tenders that helped me out down in

the Keys: Justin Lewis, Niki Norton, Andrea Schmidt, Chris Malinowski, Jess Cusick, Kelly

Kingon, Zach Boudreau, Mark Scott, Travis Richards, and Cheston Peterson. Extra special

iv thanks to Brian George for being there whenever I needed him, and to Brendan Biggs for showing me how to handle a gaggle of undergrads and for communicating via movie quotes when no one else would. Special thanks go out to Phil and Pam Greenman for giving us all a place to stay, and to Amy and Eric Peper for giving us a place to relax. Thanks to Fred and Darcy

Scott for helping out when the field crew grew by one.

A significant portion of this work would not have happened at all without the help of

Mike and Claudia Wilkerson and the Dodge Lake crew. Not only did they provide a safe (and cheap) place to tie up the boat and store gear, but Mike helped me immensely by showing me around Florida Bay and sharing the locations of some really amazing grouper holes. Mike was also a constant source of enthusiasm and curiosity about what I was doing that helped more than he probably knows.

Many thanks to the FSU Department of Biological Science; to the Miller/Winn lab; to the

Ecology Reading Group; to the Ecology and Evolution Reading and Discussion Group; and of course cheers to SEERDG. Thanks also to the FSU Coastal and Marine Lab and the faculty and staff there; thanks to the Coleman/Koenig lab; and thanks to the dive locker and ADP.

Finally, my biggest thanks go to my wife, Caitlin, whose patience and understanding throughout my career as a graduate student have been more than I could have hoped for or reasonably deserved.

v TABLE OF CONTENTS

List of Tables ...... viii List of Figures ...... xii Abstract ...... xviii

1. INTRODUCTION...... 1

2. ECOLOGICAL EFFECTS OF RED GROUPER ON FAUNAL COMMUNITIES OF FLORIDA BAY ...... 12

2.1 Introduction ...... 12 2.2 Methods...... 17 2.2.1 Physical Description of Solution Holes ...... 17 2.2.2 Field Survey Design: Faunal Community Observations ...... 18 2.2.3 Statistical Analysis: Faunal Community Observations...... 19 2.2.4 Experimental Design: Red Grouper Exclusion Experiment ...... 23 2.3 Results ...... 25 2.3.1 Variation in Physical Solution Hole Features ...... 25 2.3.2 Solution-hole Faunal Communities in the Presence and Absence of Red Grouper ...... 27 2.3.3 Red Grouper Exclusion Experiment ...... 29 2.4 Discussion ...... 32

3. INVESTIGATING THE INTERACTION BETWEEN RED GROUPER AND CARIBBEAN SPINY LOBSTER IN FLORIDA BAY SOLUTION HOLES...... 62

3.1 Introduction ...... 62 3.2 Methods...... 66 3.2.1 Lobster Abundance and Size Distribution in Solution Holes ...... 66 3.2.2 Estimating Interaction Strength ...... 68 3.2.3 Testing Lobster Avoidance Behavior ...... 70 3.2.4 Survival Estimate ...... 71 3.3 Results ...... 72 3.3.1 Lobster Abundance and Size Distribution in Solution Holes ...... 72 3.3.2 Estimating Interaction Strength ...... 74 3.3.3 Testing Lobster Avoidance Behavior ...... 76 3.3.4 Estimated Lobster Survival Adjacent to Solution Holes ...... 76 3.4 Discussion ...... 78

4. USING INVASIVE SPECIES TO TEST A BEHAVIORALLY-MEDIATED INDIRECT INTERACTION IN FLORIDA BAY ...... 94

4.1 Introduction ...... 94 4.2 Methods...... 100

vi 4.2.1 Study Site ...... 100 4.2.2 Red Grouper Effect Experiment ...... 101 4.2.3 Lionfish Effect Experiment...... 102 4.2.4 Statistical Analysis ...... 104 4.3 Results ...... 107 4.4 Discussion ...... 111

5. CONCLUSIONS ...... 129

APPENDICES ...... 133

A. EXTRA FIGURES AND TABLES ...... 133 B. IACUC APPROVAL LETTER ...... 141

References ...... 143

Biographical Sketch ...... 157

vii LIST OF TABLES

Table 2.1. Red Grouper occupancy in solution holes from three observational sites in Florida Bay from 2010 to 2013. Dark grey boxes (“N”) indicate no Red Grouper was present; light grey boxes (“Y”) indicate a Red Grouper was present; “ND” indicates no data was collected for that hole during that year...... 42

Table 2.2. Model selection results and description of the final “best-fit” models for linear mixed- effects modeling of each of the five biotic community metrics derived from diver surveys of Florida Bay solution holes. The “full” model included all listed factors plus solution-hole identity which was included in all models as a random variable. Results of the LRTs indicate the effect of either adding the interaction term or dropping each variable in question from the full model. Significant results indicate better model fit based on Chi-square analysis of the likelihood ratio statistics. Factors that significantly improved the model fit are shown in bold; if including the interaction term improved the model, then grouper presence (“grouper”) and “site” were not tested individually. AIC – Akaike’s Information Criterion – values are shown for each model to verify that the best fitting models had the lowest AIC values...... 46

Table 2.3. Results of a goodness-of-fit test of environmental correlates on species composition data from 99 solution-hole associated faunal communities with and without Red Grouper at three sites in Florida Bay between 2010 and 2013. Bold values indicate significant effects...... 50

Table 2.4. The 20 most influential species and corresponding influence scores based on both presence/absence (“P/A”), and raw abundance data from SIMPER analysis of 99 observations of solution hole faunal communities at three sites in Florida Bay between 2010 and 2013. Also included is the rank abundance of each species for comparison...... 52

Table 2.5. Standardized effects, calculated with Hedge’s g, of Red Grouper presence on the abundance of functional groups of fishes and motile invertebrates associated with solution holes in Florida Bay. Observational effects are based on 99 observations of non-manipulated solution- hole communities at three sites in Florida Bay between 2010 and 2013. Experimental effects are based on results from Red Grouper exclusion experiments conducted in 2011 and 2012. Functional group classifications for the fishes were based on reported diet information, or the location of individuals in relation to solution holes as observed during diver surveys. Functional group classifications for the invertebrates were based on reports of species known to be prey for Red Grouper. The functional group “cleaners” includes the fish and invertebrate species known to consume ectoparasites...... 53

Table 2.6. The 20 most influential species and corresponding influence scores based on both presence/absence (“P/A”), and raw abundance data from SIMPER analysis of faunal communities surveyed as part of Red Grouper exclusion experiments conducted in 2011 and 2012. Also included is the rank abundance of each species for comparison...... 58

viii Table 2.7. Species specific interaction strengths calculated with Paine’s Index from exclusion experiments conducted in 2011 and 2012, means ± S.E. from bootstrapping, and 95% confidence intervals generated from bootstrapped standard errors for the 28 species with non-negative PI values. Bold values indicate significant interaction strengths based on 95% CI...... 60

Table 3.1. Model selection results and description of the final “best-fit” models for linear mixed- effects model analysis of total lobster abundance and mean lobster size (CL) from lobsters counted and measured during diver surveys of Florida Bay solution holes. The “full” model included all listed factors plus solution-hole identity which was included in all models as a random variable. Results of the LRTs indicate the effect of either adding the interaction term or dropping each variable in question from the full model. Significant results indicate better model fit based on Chi-square analysis of the likelihood ratio statistics. Factors that significantly improved the model fit are shown in bold; if including the interaction term improved the model, then grouper presence (“grouper”) and “site” were not tested individually. AIC – Akaike’s Information Criterion – values are shown for each model to verify that the best fitting models had the lowest AIC values...... 84

Table 3.2. Model selection results and description of the final “best-fit” models for linear mixed- effects modeling for the abundance of each of the three lobster size classes counted during diver surveys of Florida Bay solution holes. The “full” model included all listed factors plus solution- hole identity which was included in all models as a random variable. Results of the LRTs indicate the effect of either adding the interaction term or dropping each variable in question from the full model. Significant results indicate better model fit based on Chi-square analysis of the likelihood ratio statistics. Factors that significantly improved the model fit are shown in bold; if including the interaction term improved the model, then grouper presence (“grouper”) and “site” were not tested individually. AIC – Akaike’s Information Criterion – values are shown for each model to verify that the best fitting models had the lowest AIC values...... 87

Table 4.1. Physical characteristics and treatment assignments for 18 solution holes located in Florida Bay that were used to test Red Grouper effects in 2012 and 2013. Treatment codes for 2012: “RG-” Red Grouper removed; “RG+” Red Grouper present. Treatment codes for 2013: “NP” no predators present; “LO” lionfish only, Red Grouper removed; “RG” Red Grouper only, no lionfish; “L+RG” both a lionfish and a Red Grouper were present. Holes # 2, 2A, and 3 were discovered during diver surveys in 2013, so no data is available for 2012. For all solution holes, the area and maximum depth values are shown as measured in 2013...... 117

Table 4.2. Estimated means of recruit abundance (± standard errors [SE]) and p-values resulting from pairwise t-tests on each of the six a priori contrasts based on the four predator treatment for each of the five community response variables at the end of the 6-week Red Grouper and lionfish experiment conducted in Florida Bay in 2013. The sample size for each predator treatment was 4. P-values less than 0.10 are indicated in bold; p-values less than 0.05 are indicated in italics...... 119

ix Table 4.3. Model selection results of linear mixed model analysis of the Red Grouper exclusion experiment conducted in Florida Bay in 2012. Likelihood Ratio Test (LRT) results are shown for the test of including a treatment by time interaction term; if the interaction did not significantly improve the model based on the LRT, then treatment and week were tested independently by dropping the term and testing that model against one with both factors. Test results of including a variance by treatment and autocorrelation structure in models were determined based on AIC scores as adding these structures caused models to be non-nested. Significant effects, shown in bold, are those whose inclusion improved the model fit; the final optimal model included all factors and structures in bold...... 120

Table 4.4. Model selection results of linear mixed model analysis of the Red Grouper and lionfish experiment conducted in Florida Bay in 2013. Likelihood Ratio Test (LRT) results are shown for the test of including a treatment by time interaction term; if the interaction did not significantly improve the model based on the LRT, then treatment and week were tested independently by dropping the term and testing that model against one with both factors. Test results of including a variance by treatment and autocorrelation structure in models were determined based on AIC scores as adding these structures caused models to be non-nested. Significant effects, shown in bold, are those whose inclusion improved the model fit; the final optimal model included all factors and structures in bold...... 121

Table 4.5. Relative change in mean recruit abundance by species for the 2012 Red Grouper exclusion experiment. The abundance of each species found during the final (week #4) diver surveys is shown as the Control value. The mean effect of Red Grouper is shown for each species and for the total abundance of juvenile reef fish recruits...... 124

Table 4.6. Relative change in mean recruit abundance by species for the 2013 lionfish experiment. The abundance of each species found during the final (week #6) diver surveys for the No Predator treatment is shown as the Control value. The mean effect of Red Grouper, Lionfish, both predators together, and the BMII effect are shown for each species and for the total abundance of juvenile reef fish recruits...... 126

Table A.1. Fish species observed at solution holes used for observational (OBS) and experimental (EXP) components of the study conducted in Florida Bay from 2010 – 2013. Functional group classifications were based on reported diet information for the Feeding group, and the location of individuals in relation to solution holes as observed during diver surveys for the Habitat group. BA = herbivores (consume primarily benthic algae); CL = cleaners (consume ectoparasites); INV = invertevores; PL = plantktivores; PV = piscivores; ZB = benthivores (consume primarily small benthic invertebrates); DEM = demersal fishes primarily observed inside solution holes; MILL = milling behavior, where the fish was found in the water-column above and around solution holes; TRANS = transient fishes observed visiting solution holes. .133

Table A.2. Motile macroinvertebrate species observed at solution holes used for observational (OBS) and experimental (EXP) components of the study conducted in Florida Bay from 2010 –

x 2013. Functional group classifications were based on reports of species consumed by Red Grouper (RG Diet) or as consuming ectoparasites (CL) from the literature...... 136

Table A.3. Results of PERMANOVA analysis of species composition data for observational data of faunal communities at solution holes; “treat” refers to the Red Grouper occupancy effect. Bold p-values indicate significant effects...... 137

xi LIST OF FIGURES

Figure 2.1. Map showing the approximate locations of the four field sites in southwest Florida Bay used in this study: Burnt Point, BP; Hawks Cay, HC; Seven-Mile Bridge, SM; and Wilkerson South, WS...... 41

Figure 2.2. Proportion of solution holes occupied by Red Grouper at three study sites in Florida Bay – Burnt Point, BP; Hawks Cay, HC, and Seven Mile Bridge, SM – from 2010 to 2013. Values shown above bars represent the number of solution holes surveyed at each site during each year of the study...... 43

Figure 2.3. Excavated area (A.) and maximum excavated depth (B.) of solution holes at three study sites in Florida Bay measured from 2010 – 2013 with and without Red Grouper. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, the open circles indicate outliers beyond those percentiles, and the values shown along the x-axis represent the number of solution holes represented by each box...... 44

Figure 2.4. Year-to-year change in the excavated area (A.) and maximum excavated depth (B.) of solution holes with different patterns of Red Grouper presence over consecutive years from three observational study sites in Florida Bay measured from 2010 to 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, the open circles indicate outliers beyond those percentiles, and the values shown along the x-axis represent the sample sizes for each occupancy pattern. Letters above boxes indicate similar groups based on Tukey HSD contrasts...... 45

Figure 2.5. Difference in the total abundance (A.) and species richness (B.) of motile fauna encountered at solution holes with and without Red Grouper at three sites in Florida Bay from 2010 – 2013. “EMP” = solution holes with no Red Grouper (N = 43); “RED” = solution holes with a Red Grouper (N = 54). The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles. P-values shown are based on the results from likelihood ratio tests between the best-fit model and model without Red Grouper presence...... 48

Figure 2.6. Difference in Shannon diversity (A.), Simpson’s diversity (B.), and Hill’s evenness (C.) of faunal communities at solution holes with (“RED”; N = 54) and without (“EMP”; N = 43) Red Grouper at three sites in Florida Bay from 2010 – 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles. P-values shown are based on the results from likelihood ratio tests between the best-fit model and model without Red Grouper presence...... 49

xii

Figure 2.7. NMDS ordination of 99 solution-hole associated faunal communities with (filled circles, N = 54) and without (open circles, N = 45) Red Grouper at three sites in Florida Bay between 2010 and 2013. Ellipses represent the standard deviation of all points for each factor (solid line, “RED” = Red Grouper present; dashed line, “EMP” = No Red Grouper), and vectors indicate significant effects of solution hole area and maximum excavated depth (“max.d”)...... 50

Figure 2.8. NMDS ordination of 99 solution-hole associated faunal communities surveyed across three sites (A.) and four years (B.) in Florida Bay. Ellipses represent the standard deviation of all points for each factor. Both year and site were significant factors (Pseudo-F based p < 0.001) in structuring faunal communities according to PERMANOVA analysis (R2 site = 0.0892; R2 year = 0.0462)...... 51

Figure 2.9. Total abundance of organisms counted in faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of Red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E...... 54

Figure 2.10. Species richness of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E...... 54

Figure 2.11. Shannon diversity of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E...... 55

Figure 2.12. Hill’s evenness of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E...... 55

Figure 2.13. Simpson’s diversity of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E...... 56

Figure 2.14. NMDS plots of community structure faunal communities associated with solution holes in Florida Bay before (A.) and after (B.) 4-week experiments conducted in 2011 and 2012. Red Grouper were excluded from experimental treatment sites labeled “REM” (open circles; N = 16), and left in control sites labeled “CON” (filled circles; N = 16)...... 57

Figure 2.15. Distribution of interaction strengths between Red Grouper and species associated with solution-hole faunal communities in Florida Bay calculated with Paine’s Index based on Red Grouper exclusion experiments conducted in 2011 and 2012...... 59

xiii

Figure 2.16. Magnitude of interaction strengths between Red Grouper and 28 species associated with solution-hole faunal communities in Florida Bay that had non-zero bootstrapped Paine’s Index values based on Red Grouper exclusion experiments conducted in 2011 and 2012. Observed means are indicated with “X” and bootstrapped means (filled circles) are shown with error bars ± 1 S.E...... 59

Figure 2.17. Size of removed Red Grouper versus the size of Red Grouper that recolonized exclusion treatment sites in Florida Bay during 2011 (diamonds; N = 8) and 2012 (circles; N = 13). The dashed line represents a 1:1 relationship; filled points indicate single recolonization events, and open points represent multiple recolonization events...... 61

Figure 2.18. Red Grouper versus maximum excavated depth of solution holes in Florida Bay based on captured lengths of fish during tagging and relocation efforts in 2011 (N = 17) and 2012 (N = 18). The solid line represents the relationship described by multiple linear regression...... 61

Figure 3.1. Lobster abundance at solution holes with (N = 53) and without (N = 46) Red Grouper (A.) and at three different sites (NBP = 8; NHC = 6; NSM = 13) in Florida Bay (B.) between 2010 and 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles. P-values shown are the results Chi-square likelihood ratio tests comparing a linear mixed model including the factor shown to a model without the factor shown...... 85

Figure 3.2. Lobster abundance by solution-hole area (A.) and maximum excavated depth (B.), and mean lobster size by solution-area (C.) and maximum excavated depth (D.). Lobsters were counted and measured in situ at solution holes in Florida Bay between 2010 and 2013. Red Grouper presence at the time of the survey is indicated by filled circles. Regression lines shown are results of simple linear regressions and indicate significantly positive relationships for all four metrics analyzed: abundance = 6.36 + 1.40 * area; abundance = 1.65 + 0.23 * depth; mean size = 4.05 + 0.377 * area; mean size = 2.78 + 0.0601 * depth...... 86

Figure 3.3. Lobster abundance by size class and total lobster abundance with (N = 53; grey boxes) and without (N = 46; white boxes) Red Grouper. Lobsters were counted and measured in situ at solution holes in Florida Bay in 2012 and 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles...... 88

Figure 3.4. Mean lobster abundance in solution holes with Red Grouper (filled circles; n = 15) and in solution holes where Red Grouper were excluded (open circles; n = 15). Data from experiments conducted in Florida Bay in 2011 and 2012. Error bars ± 1 S.E...... 89

xiv Figure 3.5. Mean lobster size measured as carapace length (CL) in centimeters over time in solution holes with Red Grouper (filled circles; n = 15) and solution holes where Red Grouper were excluded (open circles; n = 15). Data from experiments conducted in Florida Bay in 2011 and 2012. Error bars ± 1 S.E...... 89

Figure. 3.6. Proportion of lobsters by size class in solution holes with Red Grouper (“Control”; N = 15) and solution holes where Red Grouper were excluded (“Removal”; N = 15) as measured in Florida Bay in 2011 and 2012...... 90

Figure 3.7. Mean Paine’s Index values for lobster size classes and all lobsters (“Total”), calculated as the average change in abundance of lobsters in solution holes with (n = 15) and without (n = 15) Red Grouper in 2011 and 2012. Points are bootstrapped means; error bars represent 95% confidence intervals based on bootstrap results. Axes are scaled to better show error bars...... 90

Figure 3.8. Relationship between Red Grouper size and the mean size of all Caribbean Spiny Lobsters in solution holes when Red Grouper were removed in 2011 (circles; n = 9), 2012 (diamonds; n = 9), and 2013 (triangles; n = 8). All three years were pooled for analysis of the linear trend: mean lobster size = 0.0367 (Red Grouper size) + 5.3441...... 91

Figure 3.9. Abundance (A.) and mean size (B.) of lobsters collected from casitas after 48 hours with (filled points) and without (open points) a Red Grouper inside a cage attached to the top of the casita. Results shown are means ± SE from six independent trials conducted in 2012 and 2013. P-values shown are results from two-tailed t-tests...... 91

Figure 3.10. Proportion of lobsters by size class at casitas with and without a Red Grouper in a cage attached to the top of the casita after 48 hours. Proportions are pooled results from 6 independent trials conducted in 2012 (n = 2) and 2013 (n = 4)...... 92

Figure 3.11. Survival of lobsters tethered adjacent to solution holes occupied by Red Grouper after 12 and 24 hours. Stars indicate groups that were significantly different (p < 0.05) from other size classes in same time period according to pairwise χ2 test with Bonferroni corrections...... 92

Figure 3.12. Survival curve for lobsters tethered adjacent to Red Grouper occupied solution holes in Florida Bay in 2013. Points represent binary results of each tethering trial (n = 56). The curve shown is the best-fit logistical regression of the tethering results. The dashed line represents the 50% survival probability at 4.96-cm CL...... 93

Figure 4.1. Abundance of juvenile coral reef fish recruits over time at solution holes in Florida Bay with (closed circles; N = 5) and without (open circles; N = 5) Red Grouper. Values presented are means ± SE for each group. Data was collected during diver surveys of solution hole fish communities during the summer of 2012...... 118

xv Figure 4.2. Abundance of juvenile coral reef fish recruits over time at solution holes in Florida Bay from four experimental predator treatments in 2013. Values presented are means ± SE for each group; the sample size for each group was 4. Letters at the far right indicate results of pairwise comparisons performed on recruit abundance among treatments at the final census (matching letters indicate a p-value > 0.05 based on simultaneous tests performed on the best fit linear mixed-effects model)...... 118

Figure. 4.3. Species richness (A), diversity (Shannon [B] and Simpson [C]), and Hill’s evenness (D) after four weeks with (closed circles; N = 5) or without (open circles; N = 5) Red Grouper in 2012 solution hole recruitment experiment. P-values shown are results of t-tests performed on week-4 data...... 122

Figure. 4.4. Species richness (A.), Shannon (B.) and Simpson’s diversity (C.), and Hill’s evenness (D.) of juvenile reef fish recruits after six weeks with Red Grouper alone (closed circles; N = 4), Lionfish alone (triangles; N = 4), neither Red Grouper or Lionfish (open circles; N = 4), or both Red Grouper and Lionfish (diamonds; N = 4). Data was collected in 2013 from solution holes in Florida Bay. Letters indicate significant differences between groups after six weeks based on pairwise t-tests performed on the best-fit linear mixed model...... 123

Figure 4.5. NMDS ordination of juvenile reef fish communities associated with experimental solution holes with (N = 5; closed circles), and without (N = 5; open circles) Red Grouper after 4-weeks of Red Grouper presence (“Control (RG+)”) or when Red Grouper were experimentally removed from solution holes (“Removal (RG-)”). Ellipses represent the standard deviation of all points for each group...... 125

Figure 4.6. NMDS ordination of juvenile reef fish communities (abundance by species) associated with experimental solution holes at the end of the 6-week experiment conducted in Florida Bay during June – July, 2013. Ellipses show the standard deviation of all points for each predator treatment group: both Lionfish and Red Grouper (closed circles; N = 4); neither lionfish nor Red Grouper (open circles; N = 4); Lionfish Only (triangles; N = 4); and Red Grouper Only (diamonds; N = 4). There is no ellipse for the Lionfish Only predator treatment as 3 of the 4 juvenile fish communities were non-existent (n = 0) at the end of the experiment; these sites are represented by the single point located at (-1.33, 0.0217)...... 127

Figure 4.7. Bootstrapped mean effect sizes with 95% confidence intervals for the consumptive effect of a single lionfish (CELion), the total indirect interactions of a single Red Grouper (TIIRG) in 2012 and 2013, and the estimated behaviorally-mediated indirect interaction (BMII) between Red Grouper and lionfish in terms of the abundance of solution-hole associated juvenile reef fish recruits...... 128

Figure A.1. Plot of SIMPER analysis results of whole community data (closed circles, “W.C.”) and presence/absence data (open circles, “P/A”) against rank abundance from diver observations...... 137

xvi

Figure A.2. Undergraduate intern measuring lobsters in situ at a solution hole in Florida Bay, June 2011...... 138

Figure A.3. Images of casitas built to test lobster avoidance of Red Grouper. Clockwise from top left: casita frame during construction in 2012; detail of casita frame during construction showing placement of rebar supports; casita with cage and Red Grouper; casita in 2013 after one year on the bottom...... 139

Figure A.4. Evidence of triggerfish predation events on tethered lobsters in 2013...... 140

xvii ABSTRACT

Ecosystem engineers, organisms that modify the availability of abiotic resources for other species, can have complex effects on communities through a variety of direct and indirect

pathways. Describing these effects is a necessary step in understanding and predicting the effects

of engineer species. Red Grouper (Epinephelus morio) manipulate habitats by excavating

sediment and detritus from karst solution holes in Florida Bay; they are also predators that

consume a variety of crustaceans and benthic fish prey. As both a habitat manipulator and

predator, the effects of Red Grouper on communities associated with the habitats they modify

will likely be complex and difficult to predict. Here I present the results of observational and

experimental work investigating the community and species-level effects of Red Grouper on

faunal communities associated with Florida Bay solution holes.

By measuring solution holes and conducting diver surveys of faunal communities

associated with a fixed set of solution holes over time, I was able to show how the communities

and the solution holes themselves were different in the presence and absence of Red Grouper. In

general, solution holes were deeper, and communities were both more abundant and diverse

when Red Grouper were present. However, while many species changed in abundance and

occurrence from year-to-year, short term experiments, on the order of weeks, indicated that these

effects were primarily driven by a small set of species that interacted strongly with the Red

Grouper.

Caribbean Spiny Lobster (Panulirus argus) was a common and abundant member of the

solution-hole communities, and were more abundant in solution holes with Red Grouper, despite

the fact that Red Grouper are known lobster predators. Investigating the Red Grouper-lobster

interaction across lobster ontogeny indicated that Red Grouper positively affected the abundance

xviii of large lobsters, but negatively affected the abundance of smaller lobsters. To determine if this

result was due to predation or avoidance behavior by juvenile lobsters, I conducted an experiment to test relative predation risk of lobsters adjacent to solution holes with Red Grouper, and another experiment at artificial dens to test lobster avoidance in the presence and absence of

Red Grouper. The results of both experiments confirmed the hypothesis that size-selective predation by Red Grouper maintains the observed distribution of lobsters in solution holes, and not avoidance behaviors by lobsters.

The community level analysis also showed a strong positive effect of Red Grouper on the abundance of juvenile coral reef fishes. I hypothesized that this effect was caused by a

behaviorally-mediated indirect interaction (BMII) between Red Grouper and juvenile reef fishes

via solution-hole associated piscivores. A recent addition to the suite of solution-hole associated

piscivores is the Indo-Pacific lionfish (Pterois volitans and P. miles). As resident piscivores with

high site fidelity in relatively high abundance in the bay, lionfish served as a focal species to test

the BMII hypothesis. Juvenile reef fishes were most abundant at solution holes with Red

Grouper present, least abundant in solution holes with lionfish present, and at intermediate

abundances when both predators were present. These results confirmed the strong negative

effects that lionfish have on native reef fish communities, and also suggest that Red Grouper can

ameliorate some of these negative effects through a BMII that results in higher native fish

abundance.

Overall, Red Grouper had largely positive effects on the abundance and diversity of the

fish and crustacean communities found inside and around solution holes in Florida Bay. The

results presented here suggest that Red Grouper have strong effects on a wide range of species

which together shape the ecosystems they manipulate.

xix CHAPTER ONE

INTRODUCTION

Ecologists have long recognized that some organisms have a disproportionate influence

on the biodiversity of the ecosystem within which they occur. Strongly interacting organisms that

primarily act as predators are generally called keystone species, a term introduced by Paine in his

classic 1966 study and later defined as those organisms that strongly affect “the integrity or

persistence” of a community (Paine 1969). Power et al. (1996) refined the definition of a

keystone species as one “whose impact on a community or ecosystem is large, and

disproportionally large relative to its abundance.” Classic examples of keystone species are the predatory sea star Pisaster ochraceus in the rocky intertidal zone of Washington (Paine 1966) and the sea otter Enhydra lutris in the northeastern Pacific (Estes and Palmisano 1974). In each case the removal of the predator, either by experimental design or by accident, resulted in modifications to the biotic community beyond what would be predicted based solely on the abundance of the predator. In the case of P. ochraceus, consuming the competitively dominant mussel, Mytilus californianus, modified the limiting resource – settlement space – allowing for a

more species rich assemblage of intertidal settlers (Paine 1966; 1969). Importantly, similar

effects were not observed by modifying the abundance of other predators (e.g. carnivorous

gastropods), thus indicating the “keystone” role of P. ochraceus in this system. In the case of E.

lutris, predation by the otter on the sea urchins Strongylocentrolas polyacanthus and S.

franciscanus limited urchin herbivory on the macroalgae that create important shelter, substrate,

and food sources for the species rich “kelp forest” community (McLean 1962; Estes and

Palmisano 1974). On Aleutian Islands where sea otter density was low or non-existent, urchins

were highly abundant and the benthic substrate denuded of macroalgae; however, on islands

1 where sea otter populations had recovered, urchin populations were much smaller and

macroalgal forests were present (Estes et al. 1978). Both of these examples highlight the ability

of predators to have large effects on biodiversity by consuming a competitively dominant prey.

Another group of strongly interacting species, commonly known as ecosystem engineers, are those that “directly or indirectly modulate the availability of resources to other species”

(Jones et al. 1994). Classic examples of ecosystem engineers include the beavers (Castor

canadaensis [Wright et al. 2002] and C. fiber [Rosell et al. 2005]), prairie dogs (Cynomys

ludovicianus [Ceballos et al. 1999] and C. gunnisoni [Bangert and Slobodchikoff 2002]), and the

gopher tortoise (Gopherus polyphemus [Diemer 1986]). Ecosystem engineers primarily affect

ecosystem biodiversity by modifying the amount or accessibility of abiotic resources within

ecosystems, in contrast to keystone species whose effects often indirectly include changes to

resources. Modification of resources by engineers causes changes to both the composition of the

biotic community and the physical appearance of the ecosystem. For example, beavers in North

America and Europe build dams that block streams to form ponds. Pond formation causes shifts

in plant and invertebrate communities from communities composed of organisms that favor

running water to those that favor still water, including some plant species consumed by the

beavers (Naiman et al. 1998). Interestingly, the species richness of pond communities is

essentially the same as the stream communities; however across landscapes that have beaver

ponds of different ages, there is an overall increase in species richness (Wright et al. 2002).

Similar differences in patch-scale versus landscape-scale diversity have been found in numerous

other studies of ecosystem engineers (see review by Wright and Jones 2004).

Both prairie dogs and gopher tortoises are “diggers” that modify ecosystems by digging

complex underground tunnel systems. Prairie dogs have been shown to increase the diversity of

2 both arthropods and small mammals that reside within burrow structures (Ceballos et al. 1999;

Bangert and Slobodchikoff 2002). Similar to beavers, the species richness of burrow

communities does not generally differ significantly from surrounding grassland habitats; another

example of similar patch level diversity between engineered and non-engineered habitats.

However, due to differences in species composition between burrow and grassland communities, prairies with prairie dogs have higher landscape-level diversity than prairies without prairie dogs

(Miller et al. 1994).

The gopher tortoise, a digger native to xeric habitats in the southeastern US, excavates large burrows that function as shelter for at least 60 species of vertebrate and 300 species of invertebrate animals (Jackson and Milstrey 1989). Some burrow inhabitants create their own modifications of gopher tortoise burrows, creating a “burrowing cascade” which may last for multiple years following abandonment of the main burrow by the tortoise (Kinlaw and

Grasmueck 2012). Burrowing activity creates soil mounds that differ significantly in nutrient composition compared to the surrounding topsoil, resulting in mound-associated plant communities that are different from surrounding habitats (Kaczor and Hartnett 1990). Like beavers and prairie dogs, the presence of gopher tortoises in forest ecosystems results in higher landscape-level species diversity due to the succession of active and abandoned burrows. Unlike the previous examples, however, gopher tortoise activity also results in greater diversity at the patch as well, an effect that is likely due in part to the fact that some burrow resident species are both endemic and obligate commensals with gopher tortoises (Catano and Stout 2015).

While keystone predators and ecosystem engineers may have superficially similar effects on ecosystem diversity, the scale of these effects and the mechanisms by which they affect such change are fundamentally different. Both keystone and engineer species regulate the membership

3 and abundance of species in local communities through classic interaction pathways such as

competition, predation, parasitism, and mutualism (Gilpin and Diamond 1984). The essential

difference between keystone and engineer species is the nature of these interactions: keystone

predators affect communities mainly via biotic interactions while the effects of ecosystem

engineers on biotic communities are mediated through abiotic pathways. Often the effects of both keystone predators and ecosystem engineers are measured in terms of modifications to biotic communities, typically in the currency of diversity or species richness. Comprehensive studies of keystone or engineer species effect must therefore not only measure biotic change, but must also investigate the specific pathways by which such changes occur.

A framework for understanding the consequences of ecosystem engineering described by

Jones et al. (2010) includes four causal relationships: structural change caused by engineering activity; abiotic change resulting from structural changes to the habitat; biotic changes resulting from the structural and abiotic changes; and, finally, feedbacks to the engineer from the altered structural, abiotic, and biotic systems. When structural and abiotic changes caused by engineering result in a new or altered biotic community, the engineer then interacts directly with the individuals of the altered community. Thus engineers interact with biotic communities both directly and indirectly: indirect interactions mediated through abiotic change, and direct interactions with members of altered biotic communities that form as a result of engineering.

This conceptual model of ecosystem engineering leads to two important hypotheses: (1) these feedbacks to the engineer should be mostly positive; and (2) following the removal of the engineer the biotic community should revert to the pre-engineered state.

The first hypothesis, that engineered habitats should have positive effects on engineers, makes intuitive sense given that engineered habitats are in part structured by the activities of the

4 engineer, thus engineered habitats should benefit the engineer more than non-engineered habitats. Without any positive feedback, it may be unlikely that engineering behaviors would evolve at all. An example of positive effects for engineers following engineering is oyster reefs created by the eastern or Atlantic oyster, Crassostrea virginica. Oysters are considered an important foundation species, or autogenic ecosystem engineer (one that modifies habitats via their own physical structures, sensu Jones et al. [1994]), that grow collectively to form reefs that provide benthic habitats for a variety of fishes, crustaceans, polychaetes, and mollusks. Oysters are eaten by, among other things, mud crabs that are in turn prey for toadfish (Opsanus tau).

Grabowski (2004) found that modifying the complexity of oyster reefs by increasing the ratio of live oysters to dead shells, increased the amount of refuges available to mud crabs, which should have reduced oyster survival. However, increased habitat complexity also reduced mud crab foraging efficiency, ultimately resulting in higher oyster survival even in the absence of toadfish predators. This type of feedback, where organisms manipulate abiotic resources and accrue indirect benefits, may be more prevalent in natural systems than currently recognized (Boogert et al. 2006).

The second conclusion, that the loss of the engineer should be followed by the habitat eventually reverting to the non-engineered state, also makes intuitive sense. This concept of the durability of engineered habitat has been an important aspect of the refining of the ecosystem engineering term since its initial definition (Jones et al. 1997). The sea otter example exemplifies this concept: sea otters act as keystone predators by consuming urchins which feed on kelps, another important autogeneic ecosystem engineer. Sea otters reduce urchin herbivory on kelp, but the removal of sea otters from the ecosystem triggers a trophic cascade, resulting in the reduction in the engineer population that in turn causes a regime shift from kelp forest to urchin

5 barren (Simenstad et al. 1978). However, when sea otter populations recover, the ecosystem can shift back to kelp-dominated; in fact, kelp cover was found to be a better predictor of sea otter recovery than standard population assessments (Estes et al. 2010). Jones and colleagues (1997) predicted that such coupled engineer-keystone-predator-driven trophic cascades may actually be relatively common but difficult to recognize because the engineer may not always be part of the cascade.

To date, most studies of ecosystem engineers have focused on the first three aspects of this framework for understanding ecosystem engineering by describing the physical modifications to landscapes attributable to engineer activities and the resulting changes to associated biotic communities. These studies tend to focus on how engineering activities of species modifies associated biotic communities (e.g. burrowing ghost shrimp interactions with seagrass infauna [Burkenbusch et al. 2007]), a topic of particular interest for invasive engineers

(e.g. rocky intertidal acidians [Castilla et al. 2004]). Often the direct and indirect effects of engineering are combined to describe the overall effect of the presence of an engineer on an ecosystem, as in the case of Wright (2009), who measured the effect of beavers on species richness at landscape scale. Other studies have shown that engineer species can have indirect effects on other species that do not result from engineering activities. For example, a species of coral reef damselfish (Stegastes nigricans) that farms turf algae and defends algal patches from grazers has complex indirect effects on coral growth (White and O’Donnell 2010). Territorial behaviors by these damselfish are non-discriminatory and limit the access of both herbivores and coral predators to turf patches, resulting in both lower massive coral (e.g. Porties spp.) survival due to overgrowth by turf algae, and increased branching coral (e.g. Acropora spp.) survival due to reduced mortality from predation. A previous study found that S. nigricans territories were

6 primarily associated with branching corals, suggesting that while the first result is a direct result

of “farming” associated behaviors, the later result is an indirect effect of engineering by the

damselfish (Jones et al. 2006). This is also an example of how positive effects can feed back to the engineer indirectly from engineering activity. Few studies have explicitly investigated such feedbacks to engineer species. However, in order to fully predict the outcomes of engineering on communities, it is critical to know how engineers and keystone predators interact with the species that are affected by their activities by studying not only the effects of engineering on communities, but also investigating the direct interactions that engineers have with communities.

In some cases, a single species may act as both keystone predator and ecosystem engineer. A potential candidate “keystone engineer” is the Red Grouper, Epinephelus morio. Red

Grouper, a meso-predatory serranid native to the western Atlantic ranging from Brazil to North

Carolina, are commonly found on hardbottom habitats throughout the continental shelves of the western Atlantic (Heemstra and Randall 1993; Coleman et al. 2011). Red Grouper actively excavate habitats by scooping mouthfuls of sediment, dead algae, crustacean molts, and other detritus and depositing them outside of holes (Coleman et al. 2010). Excavation activities cause structural change to habitat features by increasing the amount of heterogeneous benthic habitat available to other organisms. Excavated “grouper holes” are colonized by a suite of fishes and motile invertebrates, some of which inevitably interact directly with the Red Grouper. Red

Grouper feed primarily on crustaceans and a number of demersal fishes (Moe 1969; Bullock and

Smith 1991).

The focal habitat of this dissertation is the karst hardbottom habitat of southwestern

Florida Bay north of Marathon, FL, which serves as important habitat for Red Grouper. Florida

Bay is a large (~ 2,000 km2) open embayment located immediately south of the Florida peninsula

7 and bounded by the Florida Everglades to the north, the Florida Keys to the south and east, and

the Gulf of Mexico to the west. Parts of Florida Bay are included within the boundaries of the

Everglades National Park and the Florida Keys National Marine Sanctuary. Florida Bay is

primarily dominated by shallow seagrass beds, but mangrove islands and shallow mud flats are

more common in the northern portion of the bay. Seagrass beds are primarily composed of

Thalassia testudinum, but also include Halodule wrightii and Syringodium filiforme, and compose anywhere from 75.4% to 95% of the bay bottom (Zieman et al. 1989; Fourqurean et al.

2002). Interspersed within seagrass beds are areas of exposed karst hardbottom and solution holes, areas where freshwater has dissolved the limestone leaving pits or holes in the bottom.

Sediments covering karst hardbottom and inside solution holes in Florida Bay vary in origin depending on their location and depth, and sediments from solution holes located in the southwestern part of the bay were primarily biogenic in origin and comprised of over 95%

CaCO3 (Coleman et al. 2010).

The excavating activities of Red Grouper in Florida Bay were documented and

experimentally verified by Coleman et al. (2010). In the same study, the authors also found

evidence that the species richness of fish communities associated with Red Grouper occupied

solution holes was higher than compared to control sites, a result that suggests that Red Grouper

may act as ecosystem engineers in this ecosystem. They did not investigate the specific physical

changes to solution-hole features, or the trophic interactions between Red Grouper and members

of the biotic communities that form around excavated solution holes. These interactions will

include both direct and indirect interactions: Red Grouper diets include some species commonly

found in Florida Bay solution holes, including Caribbean Spiny Lobster, Panulirus argus, majid

crabs, Mithrax spp., and toadfish, Opsanus spp. (Bullock and Smith 1991).

8 In this dissertation, I describe how Red Grouper alter the structural characteristics of

Florida Bay solution-holes and their associated biotic communities, and investigate some of the interactions between solution-hole colonizers and Red Grouper. Specifically, I examine: (1) how the physical features and biotic communities associated with solution holes in Florida Bay vary over time and space; (2) how patterns of diversity and abundance of faunal communities vary in the presence and absence of Red Grouper; and (3) how the interactions between Red Grouper and species that colonize solution holes shape solution-hole associated communities and patterns of diversity in Florida Bay.

My first objective was to describe the community level effects of Red Grouper on the diversity of solution-hole biotic communities. In Chapter Two, I describe the physical changes in a set of solution holes over time (2010 – 2013) by repeatedly measuring the size of each solution hole, the presence of Red Grouper, and the fish and macro-invertebrate communities that formed within and around each solution hole. Not all solution holes were occupied by Red Grouper each year, allowing me to compare the same holes with and without Red Grouper across years. In addition, I conducted a shorter term manipulative experiment to test the effect of Red Grouper presence on solution-hole associated communities within years. By combining these short and long-term observations, I can describe the cumulative effects of Red Grouper presence – that is those effects that are mediated via changes in habitat and those that result from direct interactions with Red Grouper.

In Chapter Two, I describe strong positive interactions between Red Grouper and two groups of species: Caribbean Spiny Lobsters, and juvenile coral reef fishes. In Chapter Three, I describe a set of experiments conducted to better understand the complex interactions that occur between Red Grouper and Caribbean Spiny Lobster in solution holes. Caribbean Spiny Lobsters

9 are nocturnal foragers that seek shelter inside crevice shelters during the day, and solution holes

represent an important source of diurnal crevice shelter for lobsters in Florida Bay (Butler et al.

1995; Herrnkind et al. 1997). Lobsters were often encountered in high abundance in solution holes with and without Red Grouper, despite the fact that lobsters are also preyed upon by Red

Grouper. Because lobsters are vulnerable to predation by Red Grouper when they are small

(Schratwieser 1999), I investigated the point at which predation risk no longer limited co- habitation with their predator. This was done through a series of experiments that measured the effect of Red Grouper presence on various life stages of Caribbean Spiny Lobsters, estimated survival of tethered individuals near solution holes with and without Red Grouper, and tested shelter selection by smaller lobsters. Together these experiments reveal how the interaction between Red Grouper and Caribbean Spiny Lobsters changes across lobster ontogeny.

In Chapter Four, I describe the interaction between Red Grouper and the suite of juvenile coral reef fishes that settle from the plankton to hardbottom habitats in Florida Bay, including the grunts (Family Haemulidae), angelfishes (Family Pomacanthidae), and parrotfishes (Genus

Scarus). I hypothesized that behavioral interactions between Red Grouper and solution-hole associated predators substantially reduced predation risk on these small fishes. The invasion of

Florida Bay by the Indo-Pacific Lionfish (Pterois volitans) provided the perfect opportunity for testing this hypothesis. Lionfish, relatively sedentary and highly effective lie-and-wait predators have significantly reduced native fish populations on reefs and hardbottom habitats throughout the western Atlantic where they have invaded (see review by Côté et al. 2013). I could easily manipulate the presence of both Lionfish and Red Grouper at solution holes to test this hypothesis, while also providing important information about the ability of native fish populations to ameliorate the negative effects of the Lionfish invasion.

10 Collectively, the observations and experiments that I present here provide essential

information on the structural effects that Red Grouper engineering activities have on solution

holes and the trophic and non-trophic effects they have on the biotic communities associated with those holes. Understanding the role of engineer species on the community dynamics of ecosystems is critical for maintaining ecosystem services and intact ecosystems, especially when the engineer in question is an exploited species (Coleman and Williams 2002).

11 CHAPTER TWO

ECOLOGICAL EFFECTS OF RED GROUPER ON FAUNAL COMMUNITIES OF FLORIDA BAY

2.1 Introduction

Organisms that produce large modifications to their habitats have generally positive effects on the species diversity of biotic communities associated with the modified habitats

(Jones et al. 1997; Bruno and Bertness 2001). This pattern holds true for organisms that generate habitat as a function of their presence, known as foundation species (sensu Dayton 1972) or autogenic ecosystem engineers (sensu Jones et al. 1994), and also for those organisms that mechanically modify habitats, also known as allogenic ecosystem engineers (sensu Jones et al.

1994). As defined by Jones et al. (1997), ecosystem engineering is “the physical modification, maintenance, or creation of habitats” which results in complex ecological effects on local biotic communities. The net positive effect of engineers on local species diversity occurs where engineers create habitat, allowing more species to live (Jones et al. 1997). Species diversity is strongly associated with community and ecosystem function, highlighting the important role that habitat modifiers play in ecosystems (Naeem et al. 1994; Coleman and Williams 2002;

Stachowicz et al. 2007).

The effects of organisms that mechanically modify habitats on local species diversity are mainly indirect: the engineer species modifies a habitat for its own purposes and in the process indirectly alters the availability and types of resources for other species. Many engineer species modify habitats by mechanically digging pits or burrows which are used by other species. For example, gopher tortoises (Gopherus polyphemus), native to xeric habitats in the southeastern

US, dig burrows that function as shelter for at least 60 species of vertebrate and 300 species of

12 invertebrate animals (Jackson and Milstrey 1989). Some burrow inhabitants create their own

modifications, creating a “burrowing cascade” which can last for many years, even after the

gopher tortoise has abandoned the main burrow (Kinlaw and Grasmueck 2011). In another

example, prairie dogs (Cynomys ludovicianus and C. gunnisoni) increase the diversity of arthropods (Bangert and Slobodchikoff 2002) and small mammals (Ceballos et al. 1999) that reside within burrow structures.

In marine ecosystems, predatory fish are known to generally have negative effects on the local species diversity of biotic communities (Hixon 1991; Almany and Webster 2004; Stallings

2009; Stier et al. 2014). These effects are particularly strong on young fishes that have recently settled to benthic habitats from the plankton, usually referred to as “recruits” (see review by

Almany and Webster [2006]; and more recent experimental evidence from Heinlein et al. [2010] and Albins [2012]). Though less well studied in marine systems, fish predators can also have negative effects on the abundance and diversity of benthic invertebrate prey (Keough and

Downes 1982; Stier 2014). The direct effects of fish predators in marine habitats are in turn modified by factors that alter the density of prey, such as habitat complexity (Beukers and Jones

1997; Anderson 2001), or the density of the predator, such as fishing (Hixon and Carr 1997;

Stallings 2008).

In contrast, some marine predators are known to have positive indirect effects on the local species diversity of biotic communities. Classic examples include keystone predators like the seastar, Pisaster ochraceus, and the sea otter, Enhydra lutris. In rocky intertidal systems, predation by Pisaster reduces the density of the competitively dominant mussel, Mytilus californianus, thereby reducing competition for space and allowing for a more species rich assemblage of intertidal settlers (Paine 1966; 1969). Manipulation of other predators (e.g.

13 carnivorous gastropods) did not result in similar effects on the settlement community, supporting

the “keystone” role of Pisaster in this system. In kelp forest communities in the northeastern

Pacific, predation by sea otters on urchins (Strongylocentrolas polyacanthus and S. franciscanus) limits urchin herbivory on the macroalgae that create important shelter, substrate, and food sources for the species rich “kelp forest” community (McLean 1962; Estes and Palmisano 1974;

Estes et al. 1978). In both of these examples most of the positive effects of the predator accrue via indirect species interactions between the predator and the biotic community. Removal of the predator either by experimental design or accidental means results in a significantly modified biotic community beyond what would be predicted based solely on the abundance of the predator; when keystone predators are allowed to recolonize an area, the ecosystem reverts to its initial state (Estes and Duggins 1995).

Habitat engineers themselves have both positive and negative effects on the species associated with modified habitats. For example, the beavers Castor canadensis and C. fiber are

often considered the “classic” ecosystem engineers as they build dams and convert streams into

ponds (Wright et al. 2002; Rosell et al. 2005). The formation of these ponds causes

corresponding shifts in the plant and invertebrate communities from those that favor running

water to those that favor still water (Naiman et al. 1998). Interestingly, the species richness of

pond communities is essentially the same as in stream communities but due to the low overlap in

species composition between streams and ponds, beaver presence results in higher landscape-

level species richness due to the diversity of patches (beaver ponds of differing age) across larger

spatial scales (Wright et al. 2002; Wright 2009). The direction of species interactions between

beavers and stream or pond associated organisms is dependent on the preferred habitat of each

organism.

14 The subject of this study is the Red Grouper (Epinephelus morio), a large-bodied predatory fish that excavates sediment and detritus from karst solution holes thereby increasing the amount of habitat available to itself and other organisms (Coleman et al. 2010). However, unlike gopher tortoises and prairie dogs, Red Grouper are also predators that sometimes consume some of the individuals that colonize the modified habitats. Because they act as both a habitat modifier and a predator, the cumulative effects of Red Grouper on the species diversity of biotic communities may be complex and potentially difficult to disentangle. Species associated with the habitats modified by Red Grouper will interact with Red Grouper through multiple pathways both direct and indirect, positive and negative. For example, Coleman and colleagues (2010) found that fish communities associated with Red Grouper modified karst solution holes in

Florida Bay were more abundant and species rich compared to those found in surrounding habitats. Such a pattern could result from several different facilitative species interactions, including those mediated by changes in resources (e.g. habitat availability), or behaviorally mediated indirect interactions (BMIIs), each of which can result in increased local species diversity (Hacker and Gaines 1997; Bruno and Bertness 2001; Dill et al. 2003; Stallings 2008).

The benthic habitat of Florida Bay is comprised of mostly seagrasses with occasional patches of exposed karst limestone. These exposed limestone patches are often pockmarked with solution holes, complex sub-benthic crevices that are excavated by Red Grouper and colonized by a suite of fish and invertebrate species (Coleman et al. 2010). Solution-hole habitats may be particularly important for some species like the Caribbean Spiny Lobster (Panulirus argus) that use crevice shelters as daytime refuge (Hernnkind et al. 1997). Diet studies of Red Grouper have consistently found that about 80% of stomach contents consisted of decapod crustaceans

(Randall 1965; Moe 1969; Bullock and Smith 1991; Brule et al 1993; Weaver 1996).

15 Cephalopods, amphipods, stomatopods, and demersal fishes (e.g. toadfishes, gobies, blennies, and cardinalfishes) make up the remaining 20% of Red Grouper diets (Bullock and Smith 1991).

One study that specifically investigated the diets of juvenile Red Grouper (individuals less than

40-cm total length [TL]) off the Campeche Bank found that decapod crustaceans dominated the diet, making up 81 – 93% of all stomach contents (Brule et al. 1994). A study of Red Grouper caught on the West Florida Shelf found evidence of a diet shift from 4:1 to 1:1 crustaceans to fish occurring around 50-cm TL, a trend that had been suggested previously in the literature

(Moe 1969; Weaver 1996). Some species commonly found in Florida Bay solution holes are also

Red Grouper prey, including Caribbean Spiny Lobsters (Panulirus argus), majid crabs (Mithrax spp.), and toadfish (Opsanus spp.).

The primary objective of this chapter is to investigate the cumulative ecological effects of

Red Grouper on biotic communities associated with excavated solution holes in Florida Bay.

Because Red Grouper act as both habitat engineer and predator in this system, the predicted effects on species diversity are not necessarily straightforward. There are two components to the study, with different objectives: (1) multi-year observational surveys to measure community and functional group level effects of Red Grouper presence on biotic communities, and to describe how the physical dimensions of solution holes change over time with and without Red Grouper; and (2) short-term manipulative experiments to measure the effect of Red Grouper presence at the community, functional group, and species levels. Using both observational and experimental components will allow me to estimate both the cumulative (direct plus indirect) and direct effects of Red Grouper presence on biotic communities associated with solution holes in Florida Bay.

16 2.2 Methods

2.2.1 Physical Description of Solution Holes

Between 2010 and 2013 I surveyed 55 different solution holes spread across four sites in southwestern Florida Bay (Figure 2.1): Seven-Mile Bridge (SM, N = 13), Burnt Point (BP, N =

8), and Hawks Cay (HC, N = 6) – holes studied previously by Coleman et al. (2010) – and

Wilkerson South (WS, N = 28) – a new site that was shared with me by a local fisherman in

2011. Solution holes at the SM, BP, and HC sites (N = 27) were used for the observational

component of the study, while solution holes at WS (N = 28) were used for the experimental

component. All solution holes used in the study were located in water less than 4-meters in depth

(range 2.4-m to 3.7-m depth).

To determine how solution holes varied in size over time, a team of two SCUBA divers

measured the area and the depth of solution holes. Area was determined by measuring the two

longest perpendicular distances across each solution hole and taking the product of these

dimensions. The number of openings to each solution hole was counted and the depth of all

openings was measured using a rigid, marked pipe. Maximum excavated depth was determined

as the greatest measured depth of all openings for each solution hole. Not all solution holes were

measured all years; for example, in 2010 excavated depth was only measured at SM. Overall I

recorded a total of 81 measurements of solution hole area and 72 measurements of excavated

depth at 23 of the 27 holes over the four years of the study.

To determine how Red Grouper modified physical characteristics of solution holes over

time, I calculated the change in the area and excavated depth for every pair of observations made

over consecutive years (see Table 2.1). These data were then grouped into one of four possible

combinations of Red Grouper presence for consecutive years: both years with Red Grouper

17 present (Y-Y), both years with Red Grouper absent (N-N), first year with Red Grouper present and second year Red Grouper absent (Y-N), and first year with Red Grouper absent and second year Red Grouper present (N-Y). I tested for differences in the change in solution-hole area and maximum excavated depth, and in the total abundance and species richness between the four occupancy combinations using one-way ANOVAs.

2.2.2 Field Survey Design: Faunal Community Observations

The abundance and species diversity of faunal communities associated with solution holes were surveyed annually during the summer months of 2010 to 2013 at the SM, BP, and HC sites. At each solution hole, a team of two SCUBA divers conducted a census of all fishes and motile invertebrates found within, in the water-column immediately above, and in a one-meter swath around each hole. Fishes were counted using the point-count method, where each diver remained stationary and counted all individuals observable for a minimum of five minutes.

Cryptic fish species and motile macroinvertebrates were counted by examining inside all areas of the solution hole after the point-count survey was completed. Both divers counted all species present, and the abundance of each species was determined as the maximum number of individuals observed by a single diver. Total species richness was determined as the sum of all species observed by both divers. All organisms were identified to species with two exceptions: juvenile Grunts (Haemulon spp.) that were < 5-cm TL and Peppermint Shrimp (Lysmata spp.).

For both Grunts and Peppermint Shrimp, multiple closely related species appear very similar so that positive identification at the species level was impossible during diver surveys. At least six species of larger juvenile Grunts and sub-adults were observed during the study, all of which were virtually identical to each other below > 5-cm TL. The Peppermint Shrimp species complex likewise is represented by at least five closely related species that may co-occur in Florida Bay,

18 all of which have distinguishing color patterns that were not easily distinguishable while SCUBA

diving (Rhyne and Lin 2006).

Red Grouper occupancy at a given solution-hole was determined by the positive identification of an individual within a solution hole during a diver survey. Divers took care to closely observe all parts of the solution hole to determine if a Red Grouper was present during the survey. The actual number of solution holes for which faunal communities were surveyed during each year varied somewhat. Of the 27 solution holes used in the 4-year observational study, 19 were surveyed all 4 years and 7 were surveyed 3 of the 4 years, and one hole added at the BP site in 2012 resulted in only 2 years of observations. Missing observations for the 7 sites not surveyed in all 4 years were because the solution holes could not be located during one of the survey years. While the exact timing of each field season varied somewhat throughout the study, all years of the survey overlapped mid-June at some point during the census (e.g. some years started in late May, others started early June, etc.).

2.2.3 Statistical Analysis: Faunal Community Observations

In addition to total abundance of all fishes and motile invertebrates, I calculated diversity metrics based on Hill’s numbers for each community observation. Hill’s numbers provide a means of calculating three commonly used diversity indices using the single equation:

1 S 1−a  a  H a = ∑ pi   i=1 

where pi is the relative proportion of the community made up by species i (Hill 1973). When

evaluated for integer values of a of 0 to 2, Ha reduces to species richness, the antilog of the

Shannon-Weiner index, and the reciprocal of the Simpson’s Index, respectfully. Generally, as a

19 increases, the index gives greater weight to more abundant species. I also calculated evenness using the equation

= / which is the ratio of the which is the ratio of� the �Simpson2 �1 ’s reciprocal and the Shannon-Weiner antilog. Evenness represents a measure of the distribution of the abundances of each species in a community, and Hill initially proposed this version of evenness because it does not include species richness (H0) thus is relatively insensitive to sample size (1997). This version of evenness converges on 1 when all species are equally abundant, so lower values indicate more uneven communities. Despite widespread disagreement in the literature regarding the use of evenness indices (and a tendency of authors to review all the available indices and conclude by offering yet another index of evenness [see Alatalo 1981, Smith and Wilson 1996, Hill 1997,

Chao et al. 2014]), Hill’s evenness ratio and its utility in measuring the distribution of individuals across species has been evaluated positively by a number of authors (e.g. Alatalo

1981; Peet 1974; Heip et al. 1998). I included it here because the specific ways that Red Grouper may modify faunal communities are unknown so calculating both Hill’s numbers and evenness should give a sufficient variety of metrics to evaluate changes in faunal communities that could be attributed to Red Grouper presence. Because I effectively treated Red Grouper as a treatment variable, they were not included as a species when calculating metrics.

I tested for the differences in biotic community metrics – total abundance, Hill’s numbers, and Hill’s evenness – based on Red Grouper presence, site, and solution-hole area and maximum excavated depth with linear mixed effects analysis. Abundances were square-root transformed for all analyses to conform to assumptions of normality. Red Grouper presence, site, solution-hole area, and maximum excavated depth were all treated as fixed effects in the model,

20 while solution-hole identity was included as a random effect to account for the repeated

measures nature of the study. Visual inspection of residual plots did not reveal any obvious

deviations from homoscedasticity or normality. Model selection was based on likelihood ratio

tests (LRT) of nested models which compared a full model that included all six factors against a

model without each factor. The final “best-fit” model for each community metric included only

those factors that significantly improved the model fit based on Chi-square tests of likelihood

ratio statistics. Akaike’s Information Criterion (AIC) values were also calculated and reported

for each model to confirm that the final model had the lowest AIC value. Because Red Grouper

occupancy varied among the three observational sites, I first conducted a LRT of the full model

versus the full model with a grouper by site interaction term. If including this interaction

improved the model, then grouper presence and site were not tested with individual LRTs. All

analyses were performed in R (R Core Team 2014) using the “lme4” package (Bates et al. 2012).

To detect differences in community composition, I tested for differences in community

structure attributable to Red Grouper presence, site, and year with permutational multivariate

analysis of variance, or PERMANOVA (Anderson 2001). PERMANOVA is similar to

traditional multivariate analysis of variance (MANOVA), but does not require that data conform

to multivariate normal distributions. I tested for differences in community structure attributable

to solution-hole area and maximum excavated depth with the “envfit” function in R (Okansen

2013) which is permutational method of testing for effects of unconstrained predictors on community structure. To quantify the contribution of individual species in driving differences in community structure, I used similarity percentage analysis (SIMPER). Because SIMPER may inflate the contributions of abundant species, I tested for individual species contributions to community structure using both a species abundance matrix and a presence/absence matrix

21 (Warton et al. 2012). To visualize patterns of community structure between various groups, I

used non-metric multi-dimensional scaling (NMDS; McCune and Grace 2002) to ordinate the data. All NMDS ordinations were performed using raw abundance data and the default options set by the “metaMDS” function of the “vegan” R package (Oksanen et al. 2015).

Finally, to compare the effect of Red Grouper presence on different functional groups of organisms, I calculated the standardized effect size, g, using the equation

/ = ����ℎ − �� � � ������ where mwith is the mean group abundance with Red� Grouper, mw/o is the mean group abundance

without Red Grouper, spooled is the pooled within-group variance, and J is a scaling factor to reduce sampling size bias (Hedges 1981; Nakagawa and Cuthill 2007). The use of standardized effect scores has previously been used to compare across studies that use different methodologies or that study different organisms (Osenberg et al. 1997; Nakagawa and Cuthill 2007). For this study, I wanted a way to compare the effect of Red Grouper on different functional groups, and also between the observational and experimental components of the study (described below).

Calculating standardized effect sizes using Hedge’s g requires that groups are independent and normally distributed (Hedges 1981). The assumption of independence is violated here because solution holes were grouped across years and sites based on Red Grouper presence. However, I felt justified in doing so because the mixed model results showed that only Red Grouper presence and excavated depth were significant factors for abundance (see Table 2.2). Abundance data were not normally distributed, but instead of transforming the data which may alter the interpretation of count data (see O’Hara and Kotze 2010), I used bootstrapping to nonparametrically estimate effect sizes and confidence intervals. To construct functional groups,

I separated fishes based on diet and habitat preference. Diet groups were based on reported diets

22 from Fishbase; and included two groups of predators – piscivores and invertivores (fishes known to consume mostly other fishes or mostly motile invertebrates, respectively) – plus benthivores

(fishes known to consume mostly benthic prey), planktivores, and herbivores (Froese and Pauly

2014; see Appendix A). Habitat preferences were based upon where fish species were observed relative to solution holes: species that were always found inside holes (“demersal”), species found milling about above or adjacent to holes (“water-column”), or species that visited holes temporarily were not observed milling about (“transient”). Motile invertebrates were grouped based on whether they are known prey of Red Grouper or not (see Appendix A). I also included a group of “cleaners” – organisms known to consume ectoparasites from other species – which included both fish and invertebrate species. Effect sizes and bootstrapped 95% confidence intervals were calculated using the “bootES” package in R (Gerlanc and Kirby 2013).

2.2.4 Experimental Design: Red Grouper Exclusion Experiment

I used solution holes at the WS site to test how Red Grouper presence affected solution hole faunal communities over short time scales (days to weeks) and to measure the direct interactions that occur between Red Grouper and solution hole associated fauna. In 2011, 26 of the 28 WS sites were occupied by a Red Grouper, and in 2012 23 of the 28 solution holes were occupied. I randomly selected 14 sites with Red Grouper in 2011 and 18 sites with Red Grouper in 2012 to be used as experimental treatment sites and assigned half of each to one of two treatment groups: Red Grouper present (control), and Red Grouper excluded (treatment). The total sample size for each of the treatment groups (control and exclusion) across both years was

16. Red Grouper at the treatment sites were captured using hook-and-line fishing gear or fish trap and were measured, tagged, and released in empty solution holes located more than 5-km from the WS experimental site. Treatment sites were visited once every 48 hours to ensure that no

23 other Red Groupers had moved into the hole; if a new individual was present, it was captured,

measured, tagged, and transported to a new site as above. I tested for a relationship between

solution hole size and Red Grouper size with multiple regression analysis using the size (TL) of

all captured Red Groupers and the area and maximum excavated depth of solution holes where

they were captured.

Divers surveyed the faunal communities associated with all sites once per week for four

weeks using the same methods described for the observational part of the study. Surveys of all

the control and treatment holes were generally conducted over a 48 hour period, where half of

the holes were visited on one day and the other half on the next day. Community metrics

analyzed included total abundance, Hill’s diversity, and Hill’s evenness. Analysis of community

metrics over the duration of the experiment was tested using two-way repeated measures

ANOVA on Hill’s diversity and evenness metrics. The NMDS/PERMANOVA/SIMPER

analysis described above was used to analyze differences in community structure between

treatments.

To calculate the strength of species interactions between Red Grouper and solution hole

associated fauna I used Paine’s Index (PI; Paine 1992). This interaction index calculates the per

capita interaction strength for all species using the equation

( ) = �� − �� ��� for each species i where N is the mean prey abundance��� in the presence of a predator, D is prey abundance in the absence of a predator, and Y is a measure of predator abundance (Berlow et al.

1999). Since Red Grouper are solitary (or at least were never observed together in holes during this experiment), Y = 1 for all replicates. Paine’s Index was initially derived to quantify the effect of a consumer when the absence of said consumer resulted in a monoculture of the competitive

24 dominant prey, and has since been used to quantify the distribution of per capita effects among

community members (Berlow et al. 1999). I used it here to determine the distribution of

interaction strengths within the community and to identify which species have strong interactions

with Red Grouper. I followed the methods described in Paine (1992) where PI values for each

species are estimated using experimental replicates for N and a mean value for D, then

constructing bootstrapped confidence intervals, given that the relatively small sample size used here which may underestimate the variability in D estimate. Bootstrapped 95% confidence

intervals were calculated using the “resample” package in R (Hesterberg 2015). I also calculated

standardized effect size of Red Grouper on the mean abundance of each functional group, as

described above.

2.3 Results

2.3.1 Variation in Physical Solution Hole Features

Discounting years when a specific solution hole could not be found, I made 67 pairs of

observations of solution holes over consecutive years (Table 2.1). Overall, annual occupancy of

solution holes by Red Grouper at all 28 solution holes at the three observational sites averaged

51.8%. Two-way repeated measures ANOVA showed no differences in the probability that a

solution hole was occupied among sites (F2,18 = 1.45, p = 0.262), but was marginally significant among years (F1,21 = 3.55, p = 0.0735; Figure 2.2). The interaction between sites and years was

also not significant (F2,21 = 0.397, p = 0.280). Solution holes remained occupied by Red Grouper

across consecutive years 29 times, remained unoccupied across consecutive years 18 times,

changed from occupied to unoccupied 13 times, and changed from unoccupied to occupied seven

times. Holes at BP had the most consistent occupancy patterns, switching just 4.76% of the time

25 (one of 21 total pairs of observations), while SM holes switched occupancy 43.3% of the time

(13 of 30 pairs). Red Grouper occupancy at HC switched 33.3% of the time (5 of 15 pairs).

Solution holes were significantly larger in terms of both area (χ2 (1) = 18.0, p < 0.0001)

and excavated depth (χ2 (1) = 30.3, p < 0.0001; Figure 2.3) when Red Grouper were present. The best fit model for solution hole area also included site and year, which both were significant fixed effects (site: χ2 (2) = 10.9, p = 0.00439; year: χ2 (1) = 4.12, p = 0.0424). In addition, the interaction between site and year was also significant for solution hole area (χ2 (2) = 6.19, p =

0.0452). For maximum excavated depth, the best fit model only included Red Grouper presence;

site and year were non-significant. However, the interactions between Red Grouper presence and

site (χ2 (2) = 6.45, p = 0.0398), and site and year (χ2 (2) = 23.5, p < 0.0001) were significant for maximum excavated depth.

As Red Grouper presence in solution holes changed from year to year, so did the size of solution holes. The change in solution-hole area (F3,42 = 4.23, p = 0.0106) and the change in maximum excavated depth (F3,35 = 8.04, p = 0.000332; Figure 2.4) were both significantly

different among the four possible change-in-occupancy combinations. Post-hoc analysis with

Tukey HSD showed that for solution-hole area, Y-N was not different from N-Y but was

different from both N-N and Y-Y. For maximum excavated depth, N-N was not significantly

different from any other combination, and all other contrasts were significant. In terms of the

excavated area of solution holes, the largest loss in size occurred when a Red Grouper was

present in year one and absent year two. When the opposite pattern happened, Red Grouper

absent year one and present year two, there was a very slight increase in area observed (0.0681

m2), but on average solution holes tended to get smaller over time, even when a Red Grouper was present over multiple years. When a Red Grouper occupied a hole that was unoccupied the

26 year before, on average the maximum excavated depth increased by 16.3-cm [± 2.72] (mean ±

SE). In contrast, when a solution hole went unoccupied the year after a Red Grouper was present, the maximum excavated depth decreased (the holes filled in) by 10.8-cm [± 2.16] from the

previous year. Solution holes that were either occupied or unoccupied by a Red Grouper over

consecutive years did not significantly change with respect to their excavated depth (+ 1.45-cm

[± 2.21] and + 0.00-cm [± 3.44], respectively).

2.3.2. Solution-hole Faunal Communities in the Presence and Absence of Red Grouper

Between 2010 and 2013, I made a total of 99 surveys of solution-hole faunal

communities. Of these, 53 surveys included a Red Grouper and 46 surveys did not find a Red

Grouper. Besides Red Grouper, a total of 66 other species were identified during the observational surveys: 50 species of fish and 16 species of motile macro-invertebrates (see

Appendix A). The results of the linear mixed-effects models indicated that the final “best-fit”

model differed slightly for each community metric in terms of which factors were included in the

final model (see Table 2.2 for full model results). Red Grouper presence in solution holes had a

significant effect on the total abundance and species richness of faunal communities, but did not

significantly improve the model for the two diversity metrics or evenness (Figures 2.5 and 2.6).

Including the Red Grouper presence by site interaction signficantly improved the model for

evenness, however including either this interaction or site alone as a factor did not improve the

models for the other four metrics. Maximum excavated depth was included as a factor in the

final model for four of the five community metrics; including depth did not improve the model

for evenness (see Table 2.2). Including area as a factor improved the models for abundance,

species richness, and evenness. Including year as a factor improved the models species richness

and Shannon diversity. The final model for total abundance included Red Grouper presence,

27 solution-hole area, and maximum excavated depth; for species richness (H0), Red Grouper

presence, year, solution-hole area, and maximum excavated depth were all included in the final

model; for Shannon diversity (H1), only year and maximum excavated depth included in the final model; for Simpson’s diversity (H2), only maximum excavated depth significantly improved the

model. Finally, the best fit model for Hill’s evenness (H2/H1) included the Red Grouper presence

by site interaction and solution-hole area as significant effects in the final model.

Community structure of solution-hole faunal communities were significantly different

when Red Grouper were present compared to solution holes without Red Grouper. The NMDS

plot of faunal communities with and without Red Grouper shows two clear groups (Figure 2.7).

The stress level of the final two-dimensional NMDS plot was relatively high (0.262), however so

was the sample size (N = 99) which is known to cause high stress values (Clarke 1993). The

PERMANOVA results confirmed that Red Grouper presence had a significant effect on

community structure (Pseudo-F1,98 = 13.1, p = 0.001). Excavated area was marginally correlated with community structure (R2 = 0.0849, p = 0.054), while maximum excavated depth was significantly correlated with community structure (R2 = 0.163, p = 0.005; Table 2.3). Both site and year also had significant effects on community structure (Figure 2.8; see Appendix A).

SIMPER results listing the influence scores for the 20 most influential species from both the whole community (abundance) and presence/absence data with their corresponding ranks from each analysis are shown in Table 2.4. The juvenile Grunt species complex had the largest influence scores for both analyses; this group of at least five species was also the most abundant group overall. Analysis of the whole community dataset reveled six species – 4 fishes and 2 crustaceans – accounted for 70% of the difference in community structure. In contrast, the analysis of presence/absence data showed that 18 species – 13 fishes and 5 crustaceans –

28 accounted for 70% of the difference in community structure. Species that generally were ranked

higher by the presence/absence analysis compared to the abundance analysis included the smaller

macro-invertebrates (e.g. cleaner shrimp and Channel Clinging Crabs) and demersal fishes (e.g.

Highhats and gobies). In general, when using species abundance data, SIMPER tended to rank

species’ influence according to their rank abundance scores (see Appendix A). The standardized

effect size calculations for each function group indicated significant effects of Red Grouper

presence on the abundance of all but two groups (Table 2.5). Piscivores and transient fishes both

showed small, non-significant effect sizes.

2.3.3 Red Grouper Exclusion Experiment

A total of 50 species were identified during the exclusion experiment: 41 species of fish

and 9 species of motile macro-invertebrates (see Appendix A). Based on the repeated measures

ANOVA analysis where both 2011 and 2012 were combined, Red Grouper presence had a

significant positive effect on the total abundance (F1,143 = 58.9; p = 0.0032; Figure 2.9), species

richness (F1,143= 316.9, p < 0.0001; Figure 2.10), and Shannon diversity (F1,143 = 9.47, p =

0.0152; Figure 2.11), and a significant negative effect on Hill’s evenness (F1,143 = 0.0634, p =

0.00036; Figure 2.12) at the end of the four week long experiments. Red Grouper presence did not have a significant effect on Simpson’s diversity (F1,143 = 0.903, p = 0.344; Figure 2.13). The interaction between Red Grouper presence and time was significant for species richness (F1,143 =

39.9, p = 0.0154), and marginally significant for faunal abundance (F1,143 = 22.78, p = 0.0641).

Time since the start of Red Grouper exclusion had a significant effect on faunal abundance

(F1,143 = 34.3, p = 0.0235) and a marginally significant effect on Hill’s evenness (F1,143 = 0.015,

p = 0.0777).

29 Differences in community structure before and after the 4-week exclusion experiments are shown in NMDS plots comparing control holes with Red Grouper and exclusion holes without Red Grouper (Figure 2.14). The NMDS ordination of the faunal communities at the start of the experiment shows clear overlap between all solution holes prior to treatment assignments, which was confirmed by PERMANOVA analysis (Pseudo-F1,24 = 0.737, p = 0.599). Although the NMDS plot of faunal communities after 4-weeks appears to show separation between control and exclusion communities, PERMANOVA analysis did not find a significant effect of Red

Grouper presence on faunal community structure (Pseudo-F1,30 = 1.32, p = 0.233). However, when year was included as a factor, it was a significant effect for community structure (Pseudo-

F1,30 = 8.39, p = 0.001). Once again the SIMPER results revealed different patterns when calculated using abundance versus presence/absence data. Based on abundance the five most abundant species – Grunt recruits, juvenile White and French Grunts, and Caribbean Spiny

Lobster – plus Pederson’s Cleaner Shrimp accounted for 83.1% of the difference in community structure. When based on the presence/absence data it took 23 species to account for the same amount of difference (see Table 2.7).

The distribution of interaction strengths among species varied, based on the calculations of Paine’s Index (PI) values for each species. Of the 50 species observed during the experiment,

28 species had non-zero PI values, and 22 species had interaction strengths equal to zero, indicating that their abundance was not different between treatment groups (Figure 2.15). I calculated the bootstrapped means and corresponding standard error values, and plotted these for the 28 species with non-zero PI values (Figure 2.16). Species specific interaction strengths are shown in Table 2.7. Bootstrapping results of the mean PI values indicated that seven species or groups had PI values with 95% confidence intervals that did not include zero: juvenile White

30 Grunts (4.97 ± 2.38), Porkfish (0.36 ± 0.16), Gray Angelfish (-0.17 ± 0.063), Hogfish (-0.033 ±

0.015), Peppermint Shrimp (-0.178 ± 0.069), Sand Perch (-0.211 ± 0.091), and Spotted Cleaner

Shrimp (-0.244 ± 0.079).

In 2011 and 2012 there were 21 separate events when another Red Grouper colonized one

of the exclusion solution-holes after the initial removal during the experiment (N2011 = 8; N2012 =

13). On average, empty holes were recolonized after about a week (mean days to replacement ±

SE = 6.62 ± 1.14). Red Grouper that moved into empty solution holes were significantly smaller

than the individual that was removed prior by approximately 4-cm (mean difference ± SE = -3.91

± 1.39 cm; T20 = -2.82, p = 0.0107; Figure 2.17). Of the 21 total re-colonization events, nine were multiple re-colonization events where the removed Red Grouper was either the third or fourth individual to be removed from the hole during the experiment.

Altogether between both the observational study and the experimental exclusions, I made

35 measurements of both Red Grouper size with matching estimates of solution-hole area and

maximum excavated depth – the 16 initial relocation captures at the start of experiments in 2011

and 2012, plus 15 captures from the observational study that were obtained when Red Grouper

were captured and tagged in an effort to estimate multi-year residency patterns, an effort that ultimately proved unsuccessful. Recolonization captures, described above, were not included in the multiple regression analysis to eliminate any experimental artifacts. Multiple regression on both solution-hole size measurements indicated there was a significant positive relationship between Red Grouper size and maximum excavated depth (T2,46 = 4.70, p < 0.0001), but the analysis did not support a significant relationship between Red Grouper size and solution-hole area (T2,46 = 1.08, p = 0.228; Figure 2.18).

31 2.4 Discussion

Repeated measurements of the physical features of the same solution holes over multiple years showed that solution holes with Red Grouper were larger both in terms of area and depth; holes that were unoccupied during the previous year got bigger the following year if a Red

Grouper moved in, and, conversely, got smaller if there was no Red Grouper. Deeper solution holes were, in turn, associated with more abundant and diverse biotic communities. Increased habitat complexity has well studied positive effects on the species diversity of marine fish and invertebrate communities, a pattern that was repeated here (Luckhurst and Luckhurst 1978;

Beukers and Jones 1997; Gratwicke and Speight 2005; Stier et al. 2014). By engineering habitats, Red Grouper modify the physical dimensions of the habitat thereby creating solution holes that can host more diverse biotic communities. Interestingly, I found that larger Red

Grouper were associated with deeper holes, but not larger ones (in terms of solution-hole area), and over time made solution holes deeper but not larger. Solution holes in Florida Bay are effectively fixed features, and their maximum dimensions are defined by geological processes.

Thus it makes sense that Red Grouper have an effect on depth but not area, given that depth is the feature that is manipulated by the digging activities of the fish. This result also makes sense given the predator avoidance behavior exhibited by Red Grouper, where they seek out holes that are deep enough for them to be completely hidden inside or nearly so. Multi-year observations found that solution holes with Red Grouper were larger in terms of their area and excavated depth, which could indicate preference by individual Red Grouper for colonizing larger solution holes, but more likely reflects the habitat modification that results from Red Grouper presence.

The cumulative community-level effects of Red Grouper presence on solution-hole faunal communities, as measured from both the observational and experimental components of

32 this study, generally matched the predictions for a habitat engineer (Jones et al. 1997; Bruno and

Bertness 2003). Previous research conducted on the fish communities associated with some of the same solution holes used in this study found these communities were more abundant and diverse compared to reference sites nearby (Coleman et al. 2010). In the present study, I focused only on solution holes so that I could compare across similar habitats with and without Red

Grouper. As described by Jones et al. (2010), structural and abiotic changes that occur due to habitat engineering will have variable consequences for specific members of associated biotic communities. That is, the effects of engineering depend on both species specific responses to

modified habitats and on species specific interactions with the engineer. The NMDS and

PERMANOVA analysis of the observational data showed that over multiple years, species

assemblages with Red Grouper were structured differently compared to those found at solution

holes without Red Grouper (see Figure 9 and Table 5). In addition, Red Grouper presence had a

negative effect on the evenness of biotic communities (see Figure 8d). Evenness, as used here,

can provide another metric with which to compare the structure of species assemblages at

solution holes. Hill’s evenness (e.g. H2/H1; Hill 1973) can be thought of as the ratio of abundant

species to common species (Jost 2010). Results from the observational study indicated that there

were more common species relative to abundant ones when Red Grouper were present in

solution holes, and that species richness was higher with Red Grouper. These results are contrary

to the general effects of piscine predators that have been shown to reduce species richness and

increase the evenness of prey communities through both selective and non-selective predation

(Almany and Webster 2004; Heinlen et al. 2010). Instead, higher species richness and lower

evenness would both occur if rare and common species increased relative to abundant ones in

33 communities with Red Grouper. These results suggest that, at the community level, the

engineering effects of Red Grouper are more important than their predatory effects.

While the effects of Red Grouper on community level metrics generally aligned with

predictions of habitat engineering, the functional-group level effects of Red Grouper were more

variable. The observational data showed that the effect of Red Grouper presence on fishes was

slightly larger than for invertebrates, but in both cases the values were larger than one indicating large effects. The effect on Red Grouper invertebrate prey was slightly less than the effect on invertebrates not part of the Red Grouper diet, but again, both were significantly positive effects.

In this case it appears that both prey and non-prey alike benefited from Red Grouper presence, possibly due to increased benthic habitat available inside excavated holes. Non-prey benefited

more from Red Grouper presence likely because none of this habitat-mediated increase in abundance was lost due to predation. However, this effect is likely confounded by the fact that all Caribbean Spiny Lobsters in the study were grouped together, irrespective of size, despite the fact that adult lobsters were likely not susceptible to predation.

Red Grouper had the largest effect on planktivorous fishes and those that inhabited the water column above solution holes, both groups that were dominated by juvenile grunt recruits.

Grunts smaller than 5-cm TL generally still consume plankton and were most often observed

milling about in the water column immediately above solution holes (Cocheret de la Moriniere et

al. 2003). Once juvenile grunts reach about 5-cm TL, their diets switch to primarily small benthic crustaceans and their coloration patterns become distinguishable. Thus juvenile grunts larger than 5-cm TL, which were counted as individual species, were included in a different diet group

(“benthivores”) but remained in the “water column” habitat group.

34 Red Grouper presence had the smallest effect on other carnivores, both those that mainly

consume invertebrates and those that mainly consume fishes. The relatively small effect on other

invertivores is somewhat surprising given the strong positive effect of Red Groper on the

abundance of motile invertebrates. However, these were often also large bodied organisms (e.g.

Nurse Sharks, Goliath Grouper) that may compete with Red Grouper for space in solution holes.

The non-significant effects of Red Grouper on piscivores is, again, somewhat surprising given the large increase observed in the abundance of small juvenile fishes with Red Grouper, which

suggests that a potentially important interaction occurs between Red Grouper and piscivores.

Piscivores encountered during the study tended to be either relatively large-bodied (e.g. Black

Grouper), or were transient (e.g. Yellowtail Snapper, Bar Jacks). Reduced predation ability of

resident piscivores via competition for space or via behavioral interactions with the Red Grouper

could explain the observed increase in abundance of small juvenile fishes in solution holes with

Red Grouper.

Red Grouper presence also had a large effect on the abundance of cleaner species found

in solution holes. This group included six species – three species of fish and three species of

motile invertebrates (see Appendix A) – that are known to remove ectoparasites off of other organisms. The presence of cleaner organisms sets up mutualisms between cleaners and their clients and the potential ecological value of such symbioses has long been theorized (Limbaugh

1961). Substantive debate in the literature has surrounded the precise nature of cleaning symbioses and the importance of cleaning symbioses on driving species distributions (see Poulin and Grutter 1996, and Cote 2000). However, some more recent experimental work has shown that the presence of cleaners on reefs can positively affect local species diversity by altering habitat choices of resident fishes and visitation rates of transient fishes (Grutter et al. 2003).

35 Similar effects may occur in solution holes where the increased abundance of cleaner organisms may in turn increase colonization of holes by resident fishes. Increased visitation by transient fishes could also occur, although this was not detected in this study. The positive effect of Red

Grouper on the cleaner group represents a potentially important indirect interaction that could partly explain the observed increase in species diversity of faunal communities when Red

Grouper were present in solution holes.

The direct effects of Red Grouper on solution hole faunal communities were investigated by experimentally manipulating Red Grouper presence in solution holes for four weeks and observing changes to the community that occurred over this time period. The short duration of the experiment enabled me to focus only on the direct effects of Red Grouper on biotic community members. By assuming that indirect effects occur through changes in the habitat and occur over time scales of months to years, conducting the exclusion experiment over days to weeks allowed me to isolate the direct effects of Red Grouper. Community-level metrics followed the same pattern as the observational study, where Red Grouper presence resulted in higher abundance, species richness, Shannon diversity, and evenness of faunal communities, but had a non-significant effect on Simpson’s diversity. Time since exclusion was a significant factor for both abundance and evenness, which was likely driven by pulses of reef fish recruitment to solution holes. A study of fish recruitment in Florida Bay found peak recruitment of Grunts occurred between June and August (Thayer et al. 2000). Another study that monitored monthly fish recruitment to nearby coral reefs in the Florida Keys found generally elevated recruitment between May and November (Sponaugle et al. 2012). In 2011 and 2012 the exclusion experiment was conducted during June and July, well within the prime settlement window for fish recruitment.

36 Effects of short term Red Grouper exclusion on functional group abundance were

negligible. The only significant effect was that of Red Grouper on invertebrate predators (see

Table 2.5). However, the absolute number of invertivores encountered during the experiments

totaled just three individuals over the two years: in 2011, one Goliath Grouper and one Green

Moray; in 201, a single Green Moray. No invertebrate predators were encountered in exclusion holes in either year, so even the very low abundance of invertivores observed in the control holes resulted in a significant effect. Although Goliath Groupers and Green Morays could potentially have significant effects on invertebrate prey, given their low abundance and the relatively small effect size of Red Grouper on all invertebrate predators from the observational study, this result is likely not meaningful. The fact that no other effect sizes were significant likely indicates that the exclusion experiment was too short to detect any meaningful direct effects of Red Grouper at the functional group level, or that the number of strong direct effects of Red Grouper were limited to a narrow range of species.

In fact, only two species had significant positive interactions with Red Grouper: White

Grunts and Porkfish (see Table 2.7). Some of these individuals likely settled from the plankton to solution holes during the experiment. White Grunts and Porkfish are reef associated species and the abundance of adult grunts of any species encountered in solution holes was very low; the majority of the White Grunts and Porkfish counted here were all smaller than 15-cm TL. The

growth rate of juvenile grunts was estimated previously at 0.25 to 0.75-mm per day (Thayer et al.

2000), thus a 1-cm long grunt recruit that settled to a solution hole at the start of the experiment

would be initially counted as a juvenile grunt recruit, and later catalogued by species anywhere from 5 to 16 days after settlement. Growth of juvenile grunt recruits from the species complex into the individual species categories would help to explain the significant interaction between

37 Red Grouper presence and time since exclusion on estimates of species richness. This may also explain why the species interaction between Red Grouper and the juvenile grunt recruit complex was not significant, but species interaction between Red Grouper and larger juvenile White

Grunts was significant.

Five species had small but significantly negative interactions with Red Grouper: Hogfish,

Sand Perch, Peppermint Shrimp, Spotted Cleaner Shrimp, and Gray Angelfish. Both Hogfish and

Sand Perch are transient benthivores, which are apparently less likely to visit solution holes when those holes are occupied by a Red Grouper. Peppermint Shrimp and Spotted Cleaner

Shrimp are both members of the “cleaners” functional group, which was dominated by a third species – Pederson’s Cleaner Shrimp. Negative responses by the more rare members of this group could indicate competition for space among cleaners. Both Pederson’s and Spotted

Cleaner Shrimp are both strongly associated with anemones, which could be a limited resource around solution holes.

The angelfishes are an interesting group: although primarily considered reef associated species, both juveniles and adults of all four species commonly found in south Florida were regularly encountered in both the observational and experimental studies. Adult angelfishes primarily consume sponges, but juveniles also consume algae and cleaning behavior by juveniles of at least one species is reported in the literature (Wicksten 1995; Froese and Pauly, 2014).

Adult angelfishes are relatively large bodied, and though they were most often observed milling about above and around solution holes, they would also commonly dart into holes as divers approached. In terms of measured species interactions, three of the four species had negative interactions with Red Grouper, though only the interaction between Gray Angelfish and Red

Grouper was significant; French Angelfish were rarely encountered and had a zero interaction

38 during the exclusion experiment. In addition, Blue and Gray Angelfish had the two highest influence scores on community structure based on the presence/absence data from the exclusion experiment (see Table 2.6). Thus it may be possible that the negative interactions observed

between angelfishes and Red Grouper result from competition for space within solution holes.

However, although both small juveniles (< 10-cm TL) and adults were observed in solution

holes, they were counted together in a single category. Small juvenile angelfish may benefit from

Red Grouper presence through the same mechanisms as juvenile grunts, so further investigation

into possible ontogenetic variations in this interaction may be warranted.

A species of particular interest is the Caribbean Spiny Lobster. Red Grouper are known

predators of lobsters, thus a strong, direct interaction with Red Grouper is not entirely

unexpected (Randall 1967; Moe 1969). However the experimental evidence supports a positive

interaction, indicating that Spiny Lobsters increased in the presence of the Red Grouper, which is

somewhat at odds with the assumption of predator-prey dynamics. Spiny Lobsters are primarily

nocturnal and seek out crevice shelters during the day, and solution holes in Florida Bay are

known to host Spiny Lobsters in high abundance relative to surrounding habitats (Hernnkind et al. 1997; Bertlesen et al. 2009). Red Grouper habitat engineering likely serves to increase the availability of suitable habitat for Spiny Lobster. Because the overall interaction is positive, this suggests that habitat availability may be more important than the increased risk of mortality that comes with cohabitating with a predator. The specific nature of this interaction is investigated in more detail in Chapter 3.

Red Grouper effects on small juvenile reef fishes were particularly large according to multiple metrics analyzed here. Functional groups that included juvenile Grunts showed large effects of Red Grouper presence in the observational study. Direct estimates of species

39 interaction found significant strong positive interactions between Red Grouper and two species

of Grunts. Finally, SIMPER analysis of abundance data found that the juvenile grunt recruit complex was the most influential group on community structure in both the observational and experimental components of the study (see Tables 2.5 and 2.6). However, the exact mechanisms by which these effects manifest are unknown. Red Grouper had negligible effects on the abundance of piscivores in the community, suggesting that potential behavioral modifications to predators may drive the observed patterns. This potential BMII is further investigated in Chapter

Four.

Comparisons of both the cumulative and direct effects of Red Grouper presence on solution hole faunal communities described here show that the cumulative effects of Red

Grouper presence outweigh the direct effects. Red Grouper presence had consistently positive effects on the abundance and species diversity of faunal communities over short (days to weeks) and long (multi-year) time scales. However, strong interactions between individual species with

Red Grouper were rare, found in only 7 of the 50 interactions measured. Estimating the effects of

Red Grouper presence in a hierarchical framework (e.g. community, functional group, individual species) by using both observational and experimental methods allowed me to describe the cumulative effects of Red Grouper as both a habitat engineer and predator along with the direct effects that Red Grouper had on individual species. These results illustrate the complex effects of a predatory habitat engineer on faunal communities associated with engineered habitats.

40

Figure 2.1. Map showing the approximate locations of the four field sites in southwest Florida Bay used in this study: Burnt Point, BP; Hawks Cay, HC; Seven-Mile Bridge, SM; and Wilkerson South, WS.

41 Table 2.1. Red Grouper occupancy in solution holes from three observational sites in Florida Bay from 2010 to 2013. Dark grey boxes (“N”) indicate no Red Grouper was present; light grey boxes (“Y”) indicate a Red Grouper was present; “ND” indicates no data was collected for that hole during that year.

Site Hole # 2010 2011 2012 2013

BP 15 ND ND N N

BP 16 N N N N

BP 17 N Y Y Y

BP 18 N N N N BP 19 N N N N

BP 30 N N N N

BP 32 Y Y Y Y

BP 33 Y Y Y Y

HC 14 Y Y Y N

HC 18 Y Y N N HC 19 N N N ND

HC 22 Y N ND N

HC 24 N Y Y N

HC 27 Y Y Y Y

SM 1 N Y ND N

SM 2 N ND N N SM 3 ND Y N N

SM 4 N Y Y N

SM 5 Y N Y N

SM 6 Y Y ND Y SM 7 Y Y Y Y

SM 8 Y Y Y N

SM 9 Y N Y N

SM 9A Y Y Y N SM 10 Y Y Y Y

SM 12 N ND Y Y

SM 14 Y N Y Y

42 1.0 BP

11 11 HC 0.8 SM 6 12 6 5 0.6

7 0.4 8 8 13

7 Proportion of occupiedProportion solution holes of 5 0.2

0.0 2010 2011 2012 2013

Year

Figure 2.2. Proportion of solution holes occupied by Red Grouper at three study sites in Florida Bay – Burnt Point, BP; Hawks Cay, HC, and Seven Mile Bridge, SM – from 2010 to 2013. Values shown above bars represent the number of solution holes surveyed at each site during each year of the study.

43

Figure 2.3. Excavated area (A.) and maximum excavated depth (B.) of solution holes at three study sites in Florida Bay measured from 2010 – 2013 with and without Red Grouper. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, the open circles indicate outliers beyond those percentiles, and the values shown along the x-axis represent the number of solution holes represented by each box.

44

Figure 2.4. Year-to-year change in the excavated area (A.) and maximum excavated depth (B.) of solution holes with different patterns of Red Grouper presence over consecutive years from three observational study sites in Florida Bay measured from 2010 to 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, the open circles indicate outliers beyond those percentiles, and the values shown along the x-axis represent the sample sizes for each occupancy pattern. Letters above boxes indicate similar groups based on Tukey HSD contrasts.

45 Table 2.2. Model selection results and description of the final “best-fit” models for linear mixed- effects modeling of each of the five biotic community metrics derived from diver surveys of Florida Bay solution holes. The “full” model included all listed factors plus solution-hole identity which was included in all models as a random variable. Results of the LRTs indicate the effect of either adding the interaction term or dropping each variable in question from the full model. Significant results indicate better model fit based on Chi-square analysis of the likelihood ratio statistics. Factors that significantly improved the model fit are shown in bold; if including the interaction term improved the model, then grouper presence (“grouper”) and “site” were not tested individually. AIC – Akaike’s Information Criterion – values are shown for each model to verify that the best fitting models had the lowest AIC values.

Dependent Variable Model Chi-sq df p AIC

Abundance Full 311.47 + grouper * site 0.42 2 0.81 315.06 grouper 21.63 1 < 0.01 331.10 Final model: site 1.55 2 0.46 309.02 Abundance = year 1.73 1 0.19 311.20 grouper presence + area 10.96 1 < 0.01 320.44 area + depth + (hole) depth 35.27 1 < 0.01 344.74 Final 308.56

Species Richness (H0) Full 345.89 + grouper * site 0.41 2 0.81 349.48 grouper 7.86 1 < 0.01 351.75 Final model: site 4.11 2 0.13 346.00 H0 = grouper presence year 5.39 1 0.02 349.28 + year + area area 15.0 1 < 0.01 358.91 + depth + (hole) depth 45.3 1 < 0.01 389.22 Final 346.00

Shannon Diversity (H1) Full 310.61 grouper * site 0.27 2 0.87 314.33 grouper 0.35 1 0.55 308.96 Final model: site 1.38 2 0.51 307.99 H1 = year + depth + year 6.32 1 0.012 314.92 (hole) area 1.22 1 0.27 309.82 depth 35.44 1 < 0.01 344.05 Final 306.79

46

Table 2.2. – continued

Dependent Variable Model Chi-sq df p AIC

Simpson’s Diversity (H2) Full 285.34 + grouper * site 0.20 2 0.90 289.14 grouper < 0.01 1 0.99 283.34 Final model: site 0.84 2 0.66 282.19 H2 = depth + (hole) year 3.07 1 0.08 286.41 area 0.06 1 0.81 283.40 depth 30.47 1 < 0.01 313.81 Final 279.02

Hill’s Evenness Full -128.00 + grouper * site 6.43 2 0.04 -130.43 grouper ------Final model: site ------Evenness = year 0.11 1 0.73 -132.31 grouper * site + area 5.85 1 < 0.01 -124.90 area + (hole) depth 0 1 1 -152.62 Final -154.60

47

Figure 2.5. Difference in the total abundance (A.) and species richness (B.) of motile fauna encountered at solution holes with and without Red Grouper at three sites in Florida Bay from 2010 – 2013. “EMP” = solution holes with no Red Grouper (N = 43); “RED” = solution holes with a Red Grouper (N = 54). The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles. P-values shown are based on the results from likelihood ratio tests between the best-fit model and model without Red Grouper presence.

48

Figure 2.6. Difference in Shannon diversity (A.), Simpson’s diversity (B.), and Hill’s evenness (C.) of faunal communities at solution holes with (“RED”; N = 54) and without (“EMP”; N = 43) Red Grouper at three sites in Florida Bay from 2010 – 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles. P-values shown are based on the results from likelihood ratio tests between the best-fit model and model without Red Grouper presence.

49

Figure 2.7. NMDS ordination of 99 solution-hole associated faunal communities with (filled circles, N = 54) and without (open circles, N = 45) Red Grouper at three sites in Florida Bay between 2010 and 2013. Ellipses represent the standard deviation of all points for each factor (solid line, “RED” = Red Grouper present; dashed line, “EMP” = No Red Grouper), and vectors indicate significant effects of solution hole area and maximum excavated depth (“max.d”).

Table 2.3. Results of a goodness-of-fit test of environmental correlates on species composition data from 99 solution-hole associated faunal communities with and without Red Grouper at three sites in Florida Bay between 2010 and 2013. Bold values indicate significant effects.

Vector NMDS1 NMDS2 R2 Pr(>r)

area 0.696 0.718 0.0849 0.054

excavated depth -0.249 0.968 0.163 0.005

50

Figure 2.8. NMDS ordination of 99 solution-hole associated faunal communities surveyed across three sites (A.) and four years (B.) in Florida Bay. Ellipses represent the standard deviation of all points for each factor. Both year and site were significant factors (Pseudo- F based p < 0.001) in structuring faunal communities according to PERMANOVA analysis (R2 site = 0.0892; R2 year = 0.0462).

51 Table 2.4. The 20 most influential species and corresponding influence scores based on both presence/absence (“P/A”), and raw abundance data from SIMPER analysis of 99 observations of solution hole faunal communities at three sites in Florida Bay between 2010 and 2013. Also included is the rank abundance of each species for comparison.

Contribution, Contribution, SIMPER rank, SIMPER rank, Species Common name Rank abundance P/A abundance P/A abundance

Haemulon spp. Juvenile grunt complex 0.0464 0.2912 1 1 1 Ancylomenes pedersoni Pederson’s Cleaner Shrimp 0.0422 0.0626 2 4 4 Pareques acuminatus Highhat 0.0356 0.0195 3 7 7 Coryphopterus glaucofraenum Bridled Goby 0.0341 0.0229 4 6 8 Mithrax spinosissimus Channel Clinging Crab 0.0335 0.0164 5 10 10 Scarus iserti Striped Parrotfish 0.0269 0.0286 6 5 5 Anisotremus virginicus Porkfish 0.0266 0.0189 7 8 13 Lutjanus griseus Gray Snapper 0.0259 0.0181 8 9 12 Pomacanthus arcuatus Gray Angelfish 0.0246 0.0163 9 11 11 Lymata spp. Peppermint Shrimp 0.0238 0.0072 10 17 17 Haemulon plumieri White Grunt 0.0228 0.0640 11 3 3 Menippe mercenaria Florida Stone Crab 0.0225 0.0059 12 20 18 Panulirus argus Caribbean Spiny Lobster 0.0210 0.1264 13 2 2 Holocanthus bermudensis Blue Angelfish 0.0183 0.0116 14 14 15 Halichoeres bivittatus Slippery Dick 0.0171 0.0100 15 15 14 Gobiosoma macrodon Tiger Goby 0.0150 0.0086 16 16 21 Mycteroperca bonaci Black Grouper 0.0148 0.0049 17 22 25 Diplectrum formosum Sand Perch 0.0145 0.0055 18 21 23 Apogon binotatus Barred Cardinalfish 0.0141 0.0044 19 24 19 Stenorhynchus seticornis Yellowline Arrow Crab 0.0135 0.0067 20 19 20

52 Table 2.5. Standardized effects, calculated with Hedge’s g, of Red Grouper presence on the abundance of functional groups of fishes and motile invertebrates associated with solution holes in Florida Bay. Observational effects are based on 99 observations of non-manipulated solution- hole communities at three sites in Florida Bay between 2010 and 2013. Experimental effects are based on results from Red Grouper exclusion experiments conducted in 2011 and 2012. Functional group classifications for the fishes were based on reported diet information, or the location of individuals in relation to solution holes as observed during diver surveys. Functional group classifications for the invertebrates were based on reports of species known to be prey for Red Grouper. The functional group “cleaners” includes the fish and invertebrate species known to consume ectoparasites.

Effect size g, Effect size g, Functional group 95% CI 95% CI Observational Experimental

All Fish 1.41 1.11, 1.70 0.538 -0.281, 1.09 Herbivores 0.626 0.216, 0.943 0.224 -0.550, 1.03 Planktivores 1.02 0.792, 1.23 0.012 -0.650, 0.823 Benthivores 0.645 0.375, 0.865 0.546 -0.194, 1.07 Invertivores 0.439 0.043, 0.734 0.677 0.362, 1.11 Piscivores 0.276 -0.127, 0.716 0.151 -0.622, 0.751 Demersal fishes 0.768 0.249, 1.23 0.469 -0.215, 1.19 Water Column fishes 1.34 1.07, 1.60 0.499 -0.152, 1.10 Transient fishes 0.0239 -0.393, 0.376 0.396 -0.340, 1.04 All Invertebrates 1.12 0.682, 1.51 0.346 -0.446, 1.15 Red Grouper Prey 0.771 0.375, 1.16 0.309 -0.464, 1.10 Not Prey 0.959 0.561, 1.38 0.184 -0.479, 1.050 Cleaners 0.993 0.539, 1.39 0.395 -0.427, 1.16

53 180 CON REM 160

140

120

100

Faunal abundance 80

60 0 1 2 3 4 Time (weeks) Figure 2.9. Total abundance of organisms counted in faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of Red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E.

16

CON 14 REM

12

10 Species richness

8 0 1 2 3 4 Time (weeks)

Figure 2.10. Species richness of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E.

54

6 CON REM

5 Shannon diversity

4 0 1 2 3 4 Time (weeks)

Figure 2.11. Shannon diversity of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E.

CON 0.77 REM 0.75

0.73

0.71

Hill's Evenness 0.69

0.67

0.65 0 1 2 3 4 Time (weeks)

Figure 2.12. Hill’s evenness of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E.

55 5 CON REM

4

3 Simpson'sdiversity

2 0 1 2 3 4 Time (weeks)

Figure 2.13. Simpson’s diversity of faunal communities associated with solution holes in Florida Bay in the presence (CON; N = 16) and absence (REM; N = 16) of red Grouper. Red Grouper were experimentally excluded from the REM treatment group. Error bars are ± 1 S.E.

56

Figure 2.14. NMDS plots of community structure faunal communities associated with solution holes in Florida Bay before (A.) and after (B.) 4-week experiments conducted in 2011 and 2012. Red Grouper were excluded from experimental treatment sites labeled “REM” (open circles; N = 16), and left in control sites labeled “CON” (filled circles; N = 16).

57 Table 2.6. The 20 most influential species and corresponding influence scores based on both presence/absence (“P/A”), and raw abundance data from SIMPER analysis of faunal communities surveyed as part of Red Grouper exclusion experiments conducted in 2011 and 2012. Also included is the rank abundance of each species for comparison.

Contribution, Contribution, SIMPER rank, SIMPER rank, Common name Species Rank abundance P/A abundance P/A abundance

Blue Angelfish Holocanthus bermudensis 0.0504 0.0089 1 11 10 Gray Angelfish Pomacanthus arcuatus 0.0493 0.0104 2 10 13 Highhat Pareques acuminatus 0.0483 0.0123 3 9 9 Juvenile Grunt complex Haemulon spp. 0.0464 0.2757 4 1 3 Peppermint Shrimp Lymata spp. 0.0461 0.007 5 14 16 French Grunt Haemulon flavolineatum 0.0457 0.0718 6 4 5 Florida Stone Crab Menippe mercenaria 0.0447 0.0088 7 12 11 Hogfish Lachnolaimus maximus 0.0446 0.0062 8 16 14 Gray Snapper Lutjanus griseus 0.0441 0.0343 9 6 6 Pederson’s Cleaner Shrimp Periclimenes yucatanicus 0.0433 0.0053 10 17 19 Queen Angelfish Holocanthus ciliaris 0.0423 0.0065 11 15 15 Porkfish Anisotremus virginicus 0.0414 0.0151 12 7 7 White Grunt Haemulon plumieri 0.0409 0.2164 13 2 2 Doctorfish Acanthurus chirurgus 0.0324 0.005 14 18 20 Channel Clinging Crab Mithrax spinosissimus 0.0306 0.0133 15 8 8 Sand Perch Diplectrum formosum 0.0295 0.004 16 20 22 Stiped Parrotfish Scarus iserti 0.0288 0.0041 17 19 17 Yellowtail Snapper Ocyurus chrysurus 0.0230 0.002 18 25 25 Jackknife Fish Equetus lanceolatus 0.0213 0.0022 19 23 24 Black Grouper Mycteroperca bonaci 0.0156 0.0022 20 22 23

58

35

30

25

20

15 # species# 10

5

0 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 >1.0 Interaction Strength [PI]

Figure 2.15. Distribution of interaction strengths between Red Grouper and species associated with solution-hole faunal communities in Florida Bay calculated with Paine’s Index based on Red Grouper exclusion experiments conducted in 2011 and 2012.

7 6 5 4 3 2

Interaction Strength 1 0 -1 Species

Figure 2.16. Magnitude of interaction strengths between Red Grouper and 28 species associated with solution-hole faunal communities in Florida Bay that had non-zero bootstrapped Paine’s Index values based on Red Grouper exclusion experiments conducted in 2011 and 2012. Observed means are indicated with “X” and bootstrapped means (filled circles) are shown with error bars ± 1 S.E.

59 Table 2.7. Species specific interaction strengths calculated with Paine’s Index from exclusion experiments conducted in 2011 and 2012, means ± S.E. from bootstrapping, and 95% confidence intervals generated from bootstrapped standard errors for the 28 species with non-negative PI values. Bold values indicate significant interaction strengths based on 95% CI.

Experimental Bootstrapped effect Species Common name 95% CI effect [PI] (± S.E.)

Haemulon plumierii White Grunt 4.96 4.97 (± 2.38) 0.206, 9.737 Panulirus argus Caribbean Spiny Lobster 3.39 3.29 (± 2.57) -1.85, 8.44 Haemulom spp. Juvenile Grunt recruit 1.57 1.56 (± 1.19) -0.811, 3.95 Pederson’s Cleaner Ancylomenes pedersoni 0.520 0.512 (± 0.257) -0.0031, 1.03 Shrimp Lutjanus griseus Gray Snapper 0.482 0.482 (± 0.255) -0.0282, 0.992 Anisotremus virginicus Porkfish 0.362 0.356 (± 0.161) 0.0332, 0.679 Haemulon flavolineatum French Grunt 0.281 0.290 (± 0.449) -0.607, 1.19 Pareques acuminatus Highhat 0.0191 0.0198 (± 0.082) -0.159, 0.198 Holacanthus bermudensis Blue Angelfish -0.0125 -0.0119 (± 0.0408) -0.0945, 0.0697 Lachnolaimus maximus Hogfish -0.0333 -0.0331 (± 0.0149) -0.0629, -0.00329 Haemulon chrysargyreum Smallmouth Grunt -0.0375 -0.0368 (± 0.0362) -0.109, 0.0356 Mycteroperca bonaci Black Grouper -0.0500 -0.0515 (± 0.0485) -0.148, 0.0455 Lutjanus synagris Lane Snapper -0.0583 -0.0572 (± 0.0563) -0.170, 0.0554 Scarus coeruleus Blue Parrotfish -0.0625 -0.0624 (± 0.0607) -0.184, 0.0590 Pomacanthus paru French Angelfish -0.0625 -0.0634 (± 0.0563) -0.183, 0.0558 Balistes capriscus Gray Triggerfish -0.0625 -0.0643 (± 0.0606) -0.185, 0.0568 Haemulon sciurus Bluestriped Grunt -0.0667 -0.0685 (± 0.0438) -0.156, 0.0190 Menippe mercenaria Stone Crab -0.0750 -0.0730 (± 0.0670) -0.207, 0.0611 Scarus iserti Striped Parrotfish -0.0833 -0.0812 (± 0.0551) -0.191, 0.0290 Holacanthus ciliaris Queen Angelfish -0.100 -0.0999 (± 0.0518) -0.204, 0.00373 Acanthurus chirurgus Doctorfish -0.104 -0.103 (± 0.0686) -0.241, 0.0339 Equetus lanceolatus Jackknife Fish -0.108 -0.107 (± 0.0712) -0.250, 0.0350 Mithrax spinosissimus Channel Clinging Crab -0.145 -0.147 (± 0.0873) -0.321, 0.0279 Ocyurus chrysurus Yellowtail Snapper -0.163 -0.161 (± 0.0812) -0.323, 0.00199 Pomacanthus arcuatus Gray Angelfish -0.169 -0.172 (± 0.0632) -0.298, -0.0458 Lysmata spp. Peppermint Shrimp -0.175 -0.178 (± 0.0694) -0.317, -0.0395 Diplectrum formosum Sand Perch -0.215 -0.211 (± 0.0911) -0.393, -0.0290 Periclimines yucatanicus Spotted Cleaner Shrimp -0.244 -0.244 (± 0.0794) -0.402, -0.0848

60 60

50

40

30 Replacement (cm) size

20 20 30 40 50 60 Removed size (cm)

Figure 2.17. Size of removed Red Grouper versus the size of Red Grouper that recolonized exclusion treatment sites in Florida Bay during 2011 (diamonds; N = 8) and 2012 (circles; N = 13). The dashed line represents a 1:1 relationship; filled points indicate single recolonization events, and open points represent multiple recolonization events.

70

60

50

40

30

RG size (cm) size RG 20

10

0 0 10 20 30 40 50 60 70 80 Maximum Excavated Depth (cm)

Figure 2.18. Red Grouper versus maximum excavated depth of solution holes in Florida Bay based on captured lengths of fish during tagging and relocation efforts in 2011 (N = 17) and 2012 (N = 18). The solid line represents the relationship described by multiple linear regression.

61 CHAPTER THREE

INVESTIGATING THE INTERACTION BETWEEN RED GROUPER AND CARIBBEAN SPINY LOBSTER IN FLORIDA BAY SOLUTION HOLES

3.1 Introduction

Predation is a well-known force in structuring aquatic and marine communities (see reviews by Connell 1975; Hixon 1986; Menge 2000). Empirical evidence supports the role of predators in shaping the structure and diversity of prey communities in a variety of aquatic ecosystems including in lakes (Brooks and Dodson 1965), the rocky intertidal (Paine 1966), and

coral reefs (Connell 1978; Doherty et al. 1985). Ecological theory suggests that predator-prey

coexistence will be enhanced in habitats that are patchy and include prey refuges, those spaces

where the risk of predation is lessened (e.g. Caswell 1978; Holt 1987). The importance of habitat

complexity in regulating prey diversity has been well studied in a variety of marine ecosystems

including seagrass beds (Orth et al. 1984; Eggleston et al. 1997) and coral reefs (Shulman 1985;

Hixon and Beets 1993). When habitat quality is highly variable, the presence of predators can

force prey to occupy habitats of lesser quality, resulting in trade-offs between predation risk and

food availability (Werner et al. 1983; Heithaus and Dill 2002).

Within a prey species, predation is typically size-selective; for example, prey mortality

usually declines with increasing body size when larger prey are less vulnerable to predation

(Sogard 1997). Many marine fishes and invertebrates experience relatively high mortality during

the post-settlement period immediately following recruitment to benthic habitats from the

planktonic larval stage (Hunt and Scheibling 1997; Almany and Webster 2006). The magnitude

of early post-settlement mortality is determined by a variety of density-dependent and density-

independent factors including recruit supply (Wahle and Incze 1997; Caselle 1999; Dougherty

62 2002), predation (Carr and Hixon 1995; Steele 1997), and habitat complexity (Beukers and Jones

1997; Johnson 2007), all of which interact to regulate the adult population size and distribution

of prey species.

Caribbean Spiny Lobsters (Panulirus argus; referred to here as “lobsters”) have size- structured populations that could result in important interactions with their predators. Predators

of lobsters include a variety of piscine and invertebrate predators including groupers

(Epinephelidae), snappers (Lutjanidae), grunts (Haemulidae), triggerfish (Balistidae), toadfish

(Opsanus spp.), bonefish (Abula vulpes), permit (Trachinotus falcatus), nurse sharks

(Ginglymostoma cirratum), octopus (Octopus briareus), and portunid crabs (Randall 1965; Moe

1969; Smith and Herrnkind 1992; Mintz et al. 1994; Berger and Butler 2001; Childress and

Herrnkind 2001; Bouwma 2006). The lobster life cycle proceeds through a series of distinct, size-based benthic stages. Following a planktonic larval stage that can last up to 9 months, post-

larval or “algal-phase” lobsters settle preferentially on red algae (Laurencia spp.; Marx and

Hernnkind 1986). Algal-phase lobsters are highly cryptic, solitary, and maintain close

associations with red algae until reaching ~ 25-mm carapace length (CL; Herrnkind and Butler

1986). Post-algal phase lobsters, referred to here as “juveniles,” range in size from 26 to 45-mm

CL, and are typically found associated with sponges, coral heads, and small crevices which are

used as diurnal shelters. Sub-adult (46 to 75-mm CL) and adult lobsters (> 76-mm CL) are

highly gregarious and reside primarily in crevice shelters before mature adults migrate offshore

to spawn in the fall (Marx and Herrnkind 1986).

Ontogenetic stages in lobsters are differentiated by changes in morphology, behavior, and

habitat associations. At around 26-mm CL, lobsters undergo the most noticeable of these

changes as they change in coloration pattern and begin to aggregate in shelters with conspecifics

63 (Childress and Hernnkind 1994). Gregarious behavior begins in the juvenile stage and the strength of conspecific chemical cues on aggregated shelter use increases as lobsters grow through the sub-adult and adult stages (Ratchford and Eggleston 1998). While some authors have found that conspecific abundance in shelters was positively correlated with survival (Smith and

Hernnkind 1992; Mintz et al. 1994), others have suggested that conspecific aggregation behavior may increase individual survival by reducing the time spent searching for shelter, thus reducing time spent vulnerable to predation (e.g. the “Guide Effect” [Childress and Hernnkind 2001]).

Whatever the specific mechanism, predation on juvenile lobsters is intense, and behaviors that increase survival are likely to be under strong selection (Eggleston et al. 1990; Smith and

Herrnkind 1992; Schratweiser 1999).

In Florida Bay, the focal site for this study (see Figure 2.1), exposed areas of the karst limestone bottom are pockmarked with solution holes. These habitat features are used by juvenile, sub-adult, and adult lobsters and represent an important source of diurnal crevice shelters for lobsters in the bay (Herrnkind et al. 1997). Red Grouper, one of many lobster predators also found in Florida Bay, excavate sediment and detritus from solution holes, indirectly increasing the availability of crevice shelter habitat available to other species

(Coleman et al. 2010). Previous research suggests that survival of juvenile lobsters (< 45-mm

CL) near Red Grouper occupied solution holes may be extremely low (Schratweiser 1999).

However, the observations and experiments described in Chapter Two showed that Red Grouper had a positive effect on the total abundance of lobsters in solution holes. Furthermore, Red

Grouper presence resulted in more abundant lobster predators in solution holes, including triggerfish, Nurse Sharks, and Goliath Grouper. Conversely, some other lobster predators commonly found in the bay (e.g. octopuses, toadfish) were conspicuously absent from solution

64 holes when Red Grouper were present (Berger and Butler 2001; Chapter Two). The analysis

presented in Chapter Two considered the interaction between Red Grouper and all stages of

lobsters grouped together. However, this interaction is likely to change across lobster ontogeny.

While high predation risk around Red-Grouper-occupied solution holes has been experimentally

shown for juvenile lobsters, the interactions between Red Grouper and sub-adult and adult

lobsters have yet to be thoroughly investigated.

Lobsters use a variety of chemical cues to navigate their environment, and olfactory cues

have been shown to be important for conspecific aggregation (Ratchford and Eggleston 1998;

Butler et al. 1999), avoiding disease-infected conspecifics (Behringer and Butler 2010), and most

importantly for predator avoidance (Eggleston and Lipcius 1992; Wahle 1992; Berger and Butler

2001). Experiments that have tested predator avoidance in lobsters using mesocosms have found

that lobsters will actively avoid nurse sharks (Eggleston and Lupcius 1992), and octopuses

(Berger and Butler 2001; Butler and Lear 2009). There is little evidence to suggest that lobsters avoid teleost predators (but see Bouwma 2006); however, other lobster species (e.g. Homarus americanus) are known to use olfactory cues to avoid teleost predators (Wahle 1992). It is not yet known whether Caribbean Spiny Lobsters actively avoid Red Grouper.

The combination of ontogenetic niche shifts and size-selective predation suggest that trophic interactions between lobsters and their predators may be complex, and should be measured at the stage-level, not just at the species-level (sensu Miller and Rudolf [2011]). In this

study, I investigated the interaction between lobsters and Red Grouper in solution-hole habitats

in Florida Bay separately for each of the ontogenetic niches involved in the interaction.

Specifically, I addressed the following questions: (1) how does the lobster-Red Grouper

interaction differ in magnitude and direction across ontogenetic stages of lobster; (2) do juvenile

65 lobsters actively avoid dens occupied by Red Grouper; and (3) at what size do lobsters enter a

size-refuge from Red Grouper predation?

3.2 Methods

3.2.1 Lobster Abundance and Size Distribution in Solution Holes

From 2010 to 2013, I surveyed the abundance of lobsters residing in solution holes at

three sites in southwestern Florida Bay (see Figure 2.1). The distribution of previously identified solution holes at each site was uneven – Seven-Mile Bridge (SM, N = 12), Burnt Point (BP, N =

9), Hawks Cay (HC, N = 6) – as was the total number of holes that were surveyed during each

year. All 9 holes at Burnt Point were surveyed each year; in 2012 and 2013 one of the Hawks

Cay solution holes could not be found; all Seven-Mile Bridge holes were surveyed in 2010 and

2013, but one hole could not be found in 2011 and two holes could not be found in 2012.

In 2012 and 2013 I also measured the size of lobsters at these solution holes; lobster size

was not measured at any solution holes in 2010, and in 2011 lobster size distribution was measured, analyzed, and reported on as part of an undergraduate intern project, so those data are not included here. Lobster size, based on carapace length (CL), was estimated to the nearest centimeter in situ by a SCUBA diver with a plastic cylindrical measuring stick (see Appendix

A). This methodology was preferred over handling lobsters as a calm diver could reliably estimate the size of a large quantity of individuals efficiently and without disturbing them.

Lobsters were not handled or tagged, as this was reported to have adverse effects on lobster site preference where tagged lobsters do not return to sites where they had been handled previously

(pers. comm. Herrnkind, 2010). Lobster abundance and size estimates were taken at the same time that the team of two divers conducted a faunal census described in Chapter Two. Red

66 Grouper presence was determined at the time of the diver survey. During surveys the divers spent at least 5 minutes observing faunal communities and measuring lobsters, while closely investigating the entire solution hole with the aid of underwater flashlights. Red Grouper presence was determined by the positive identification of an individual Red Grouper within a solution hole by either one of the divers during these surveys. The exact timing of each field season varied somewhat across years, but all years overlapped mid-June.

I tested for differences in total lobster abundance, mean lobster size (CL), and abundance by size class using linear mixed effects analysis. Red Grouper presence, site, year and solution- hole area and maximum excavated depth were treated as fixed effects in the model, while solution-hole identity was treated as a random effect to account for the repeated measures nature of the study. Abundances were square-root transformed for all analyses to conform to assumptions of normality. Visual inspection of residual plots did not reveal any obvious deviations from homoscedasticity or normality. Model selection was based on likelihood ratio tests (LRT) of nested models which compared a full model that included all six factors against a model without each factor. The final “best-fit” model for each community metric included only those factors that significantly improved the model fit based on Chi-square tests of likelihood ratio statistics. Akaike’s Information Criterion (AIC) values were also calculated for each model to confirm that the final model had the lowest AIC value. Because Red Grouper occupancy varied among the three observational sites, I first conducted a LRT of the full model versus the full model with a Red Grouper presence by site interaction term. If including this interaction improved the model, then Red Grouper presence and site were not tested with individual LRTs.

When the LMM results supported including categorical factors (e.g. Red Grouper presence, site, and year) in the final model, I tested for differences between categories using one-way ANOVA

67 and Tukey post-hoc tests. Solution-hole area and maximum excavated depth were both

continuous variables, so I tested for relationships with the dependent variables using simple

linear regressions when LMM results supported including these factors in the final model. All

analyses were performed in R (R Core Team 2014), and all linear mixed-effects modeling

analysis was done with the “lme4” package (Bates et al. 2012).

3.2.2 Estimating Interaction Strength

The interaction between Red Grouper and lobsters was estimated experimentally in 2011

and 2012 at solution holes located at the WS site (see Figure 2.1). In 2011, initial diver surveys

found Red Grouper present at 26 of the 28 WS holes, and in 2012 Red Grouper were present at

23 of the 28 WS holes. I randomly selected solution holes with resident Red Grouper (N2011 = 14;

N2012 = 18) and evenly assigned holes to one of two treatment groups: Red Grouper excluded

(removal treatment) or Red Grouper present (control treatment). Red Grouper at the removal treatment holes were captured using hook-and-line fishing gear and were measured, tagged, and released at solution holes located at a similar hardbottom site located at least 5-km from the WS

experimental site. Following the initial removal of the Red Grouper, removal treatment holes

were visited once every 48 hours for the duration of the experiment to ensure that no other Red

Groupers had moved into the hole; if a new individual was present, it was captured, measured,

tagged, and released at a new site as described above.

Divers counted and measured all lobsters in each of the experimental solution holes

during the initial survey and then once per week for four weeks. Lobsters were assigned to size

classes based on ontogenetic stages: juvenile = < 5-cm CL; sub-adult = 5 to 7-cm CL; and adult

= ≥ 8-cm CL. Lobster size classes used here vary slightly from lobster stages described above as

due to the measurement method used which was only accurate to the nearest centimeter. Diver

68 surveys were conducted over a 48 hour period, weather permitting, where half of the solution

holes were surveyed on one day and the other half were surveyed the next day. Two of the

solution holes (one control and one treatment hole) were dropped from the 2011 group because

of missing data due to weather that prevented timely surveys.

I tested for changes in lobster abundance and mean size in solution holes with (control) and without (treatment) Red Grouper over time using Repeated Measures ANOVA. One-way

ANOVAs found no effect of year on either lobster abundance (F1,22 = 3.35; p = 0.0806) or mean

size (F1,14 = 0.886; p = 0.362), so data were pooled across years and grouped by treatment: Red

Grouper present (control group; N = 15) or Red Grouper excluded (N = 15). I tested for differences in the proportion of each lobster size class between control and treatment groups using one-tailed paired t-tests. The tests for juvenile lobster proportion in solution holes was one-

tailed based on the hypothesis that removal of a lobster predator should result in an increase in

the number of small lobsters in solution holes. T-tests for sub-adult and adult proportion were

both two-tailed, given that no explicit hypothesis was made for the change in the abundance of

larger lobsters.

The interaction strength between Red Grouper and lobsters was calculated using a modified version of Paine’s Index (PI; Paine 1992). This index calculates the per capita interaction strength for all species using the equation

( ) = �� − �� ��� for each lobster size class i, where N is the lobster abundance�� in the presence of the Red Grouper

(from the control treatment holes), and D is the lobster abundance in the absence of the Red

Grouper (from the removal treatment holes; Berlow et al. 1999). To calculate PI for each lobster

size class, I followed the methods described by Paine (1992) where PI values are estimated using

69 each experimental replicate for N and a mean value for D, then constructing resampled

confidence intervals, given that the relatively small sample size used here may underestimate the

variability in D parameter. Bootstrapped 95% confidence intervals were calculated using the

“resample” package in R (Hesterberg 2015).

I tested for a relationship between Red Grouper size and mean lobster size with simple

linear regression using the total length of Red Grouper that were caught for removal experiments

and the mean length of all lobsters measured at the time of the Red Grouper removal. Regression

and bootstrap analyses were conducted using the R core package.

3.2.3. Testing Lobster Avoidance Behavior

To test if juvenile lobsters avoid dens occupied by Red Grouper, I built artificial lobster

dens, also known as “casitas,” based on the design described by Eggleston et al. (1990). Casitas

were constructed by bolting 18” square concrete paving stones to a frame made out of 2”

diameter polyvinyl chloride (PVC) pipe and rebar. Two ½” x 3’ steel reinforcing bars were set in

grooves cut in the PVC pipe to provide longitudinal support and add weight to the structure (see

Appendix A). The final measurements of each casita were 91-cm wide x 137-cm long x 15-cm tall. The stacked design of the PVC frame resulted in exterior gaps approximately 5-cm tall along each side. Limiting the size of casita openings has been shown to restrict the size of both lobsters and lobster predators that can move into casitas, and the 5-cm height was chosen to restrict entrance to juveniles and smaller sub-adult lobsters (Eggleston et al. 1990). Four casitas were

constructed at the BP hardbottom site in 2012; all four casitas were separated by at least 100-m

from each other, and were at least 100-m away from known solution holes. After construction,

the casitas were left on the bottom for two weeks before the first trial was conducted.

70 For each trial, two of the four casitas were randomly selected and pre-made nylon-coated wire fish cages (approximate dimensions: 61-cm wide x 122-cm long x 61-cm tall) were placed on top of each casita (see Appendix A). Fish cages were anchored to the casita with standard concrete construction bricks, which also provided shelter for the Red Grouper. At the start of each trial all of the lobsters that were present in the two chosen casitas were captured, measured, and released at a similar hardbottom habitat at least 2-km from the BP site. Moving these juvenile lobsters to adjacent habitats was done to ensure that lobsters recruiting to the casitas during the trial had not been handled or disturbed, which was stated to have adverse effects on lobster site preference (see above; pers. comm. Herrnkind, 2010). A Red Grouper was placed in one of the two fish cages, with the other cage remaining empty as a control. After 48 hours the

Red Grouper was released alive back at the location of capture, and all the lobsters that had colonized each of the casitas with cages were captured, measured, and released back to the BP site. Trials were conducted at least one week apart. In 2012 I conducted two trials, and in 2013 I conducted four trials. Data collected from all six trials were pooled across the two years, and the abundance and mean size of lobsters in the two casitas after 48 hours were compared using two- tailed t-tests. The distributions of lobsters by size class at casitas with and without Red Grouper were also tested using two-tailed t-tests. T-tests were performed using the R core package.

3.2.4. Survival Estimate

To test how the survival of lobsters at Red Grouper occupied solution holes varied with lobster size, I conducted a tethering experiment at the WS site in 2013. Lobsters were tethered to concrete anchors placed within 1-m of Red Grouper occupied solution holes for 24 hours. Prior to tethering trials, each solution hole was surveyed by divers to determine the presence of lobster predators; tethering trials were only conducted at solution holes where Red Grouper was the only

71 lobster predator observed. Lobster tethers were constructed using a 20-cm length of 22.7-kg test

monofilament attached with a brass swivel to a 30-cm x 30-cm x 5-cm concrete paving stone.

The tether was looped over the lobster carapace between the 4th and 5th walking legs, tied with an overhand slipped loop knot, and secured to the carapace with a small amount of cyanoacrylate glue. This tethering methodology has previously been used by others to estimate survival of juvenile lobsters (Hernnkind and Butler 1986; Schratweiser 1999).

Between 1 July and 21 July, 2013, I tethered a total of 60 lobsters, 20 from each lobster size class (actual size range: 2.7 to 9.4 cm CL), adjacent to solution holes occupied by Red

Grouper. Each survival trial started in the late evening and was checked after 12 hours and 24 hours to determine if the lobster was still alive and attached to the tether. After 24 hours any

lobsters still alive were removed from the tether and released. In addition to the 60 lobsters

tethered to the concrete anchors, two more lobsters from each size class (total = 6) were tethered

to the center of a 1-m by 1-m steel frame marked with alternating 10-cm long black stripes. An

underwater camera was positioned next to the frame to record any predation events that occurred

within the first two hours of the trial. When a lobster was missing after 12 or 24 hours, the state

of the tether was noted (e.g. intact, monofilament snapped, broken swivel, or sections of

carapace remaining attached). Survival of lobsters from each of the three size classes after 12 and

after 24 hours was tested using pairwise χ2 tests with the Bonferroni correction. Tethering

experiment results were used to generate a survival curve using logistic regression. Chi-square

tests and logistic regressions were performed using the R core package.

72 3.3 Results

3.3.1. Lobster Abundance and Size Distribution in Solution Holes

Between 2010 and 2013 I made 99 surveys of lobster abundance at the three

observational sites used in the study. In 2012 and 2013 I measured a total of 397 lobsters during

these surveys. Lobster presence during surveys was variable and ranged from zero to 41 lobsters

per hole (mean ± SE of lobster abundance = 8.74 ± 0.87). The size of lobsters measured during

surveys ranged from 2-cm to 11-cm CL (mean ± SE = 5.91 ± 0.085).

Results of linear mixed model analysis indicated that Red Grouper presence, site,

solution-hole area, and maximum excavated depth of holes were all significant factors for

explaining the total lobster abundance in solution holes (Table 3.1). The abundance of lobsters

was significantly greater when Red Grouper were present during surveys (χ2 (1) = 4.55, p =

0.032; Figure 3.1a). Post-hoc analysis of site differences indicated that lobster were more abundant at the Hawks Cay site compared to the Seven-Mile Bridge site (p = 0.036), while the other contrasts were not significant (Figure 3.1b). Including both solution-hole area and maximum excavated depth improved the model (see Table 3.1), and simple linear regressions indicated that total lobster abundance was positively related to both area (F1,68 = 7.63, p = 0.007) and depth (F1,52 = 5.45, p = 0.023; Figure 3.2). For mean lobster size, the linear mixed model results supported including only solution-hole area and maximum excavated depth in the final model. Simple linear regressions indicated that both area and depth were positively related to

total lobster abundance (area: F1,30 = 4.84, p = 0.036; depth: F1,31 = 4.23, p = 0.048; Figure 3.2).

When lobster abundance was analyzed by size class, Red Grouper presence was a

significant factor in controlling the abundance of sub-adult and adult lobsters (Table 3.2; Figure

3.3). For juvenile lobster abundance the Red Grouper presence by site interaction significantly

73 improved the model (Table 3.2). Both solution-hole area and maximum excavated depth

significantly improved the model fit for the abundance of all size classes of lobsters. Regression

slopes extracted from simple linear regressions for each factor and size class indicated that

solution-hole area and depth were both positively related to sub-adult (area: 1.08; depth: 0.103)

and adult abundance (area: 0.428; depth: 0.065). Solution-hole area was negatively related to

juvenile abundance (regression slope = -0.229), while solution-hole depth was positively related

(0.005).

3.3.2. Estimating Interaction Strength

In 2011 and 2012 I made a total of 144 surveys of lobster abundance and measured lobsters during 136 surveys at the experimental site used in the study. These surveys occurred at

12 solution holes in 2011 and 18 holes in 2012, equally distributed between control and treatment groups. In total I measured 2614 lobsters across both years. No count data was missing for 2011, however 14 sets of lobster measurements made during the start of the experiment (8 measurements made during the initial diver surveys [week 0] and 6 measurements made during week 1) were excluded due to concerns regarding their validity as these were the first measurements made using the in situ methods described above. In 2012 lobster counts and measurements from initial diver surveys for 6 solution holes (3 control and 3 treatment) were missing due to foul weather that limited diving operations to visual confirmations of Red

Grouper presence via snorkel.

All of the solution holes used for this experiment were clearly utilized by lobsters, and all holes had at least 3 lobsters present during the initial diver surveys (maximum lobster abundance

= 69; mean ± SE of lobster abundance = 31.9 ± 2.80). After four weeks there were on average more and larger lobsters with Red Grouper present than absent; however, these differences were

74 not significant (abundance: Repeated Measures ANOVA F1,138 = 0.976; p = 0.325; mean size:

F1,124 = 0.189; p = 0.665). Time since the start of the experiment was a significant factor for lobster abundance (F1,138 = 8.95; p = 0.003), reflecting the observed increase in the number of

lobsters counted in both treatments over the four weeks of the experiment (Figure 3.4). However,

time since the start of the experiment was not a significant factor for mean lobster size (F1,124 =

0.738; p = 0.392; Figure 3.5), nor were the interactions between Red Grouper presence and time since the start of the experiment significant for either lobster abundance or mean lobster size.

After four weeks the distributions of lobsters in each size class were different for solution holes with and without Red Grouper (Figure 3.6). Paired t-tests showed that the difference in the proportion of juvenile lobsters was significantly greater in solution holes without Red Grouper than in solution holes with Red Grouper (T28 = -1.92; p = 0.033). However, t-tests did not show any significant differences in the proportions of sub-adult (T28 = -0.778; p = 0.443) or adult lobsters (T28 = 1.03; p = 0.312) between treatments. The interaction strength between Red

Grouper and lobsters was calculated for all lobsters using total lobster abundance and for each size class separately (Figure 3.7). Based on the bootstrapping results, the direction of the interaction was positive for all lobsters (3.67 ± 2.87), sub-adult lobsters (0.612 ± 0.23), and adult lobsters (2.23 ± 1.18), but negative for juvenile lobsters (-0.247 ± 0.10). All interactions were significant, based on the result of bootstrapped confidence intervals, none of which included zero; confidence intervals were non-symmetric and are reported in Table 3.3.

Between 2011 and 2013 I removed a total of 26 Red Grouper from solution holes for the removal experiments (see Chapters 2 and 4), and measured the size of all lobsters present in solution holes at the same time. There was a significant positive relationship between the size of the Red Grouper and the mean size of lobsters measured in the solution hole at the time of the

75 removal (F1,24 = 13.12; p = 0.001; Figure 3.8). When the year of the removal was included as a

covariate, it was not significant (F1,23 = 0.0122; p = 0.913), so all three years were pooled to

calculate the linear trend: mean lobster size = 0.037 * Red Grouper size + 5.34.

3.3.3. Testing Lobster Avoidance Behavior

Two 48-hour lobster avoidance trials were conducted at casitas in 2012, and four more

avoidance trials were conducted in 2013; each trial included two treatment casitas with fish cages

anchored on top: Red Grouper present in cage, and empty cage. After 48 hours there was no

difference in either the abundance (T10 = 1.15; p = 0.278) or the mean size of lobsters (T10 =

0.080; p = 0.938; Figure 3.9) found at casitas with or without the caged Red Grouper. There were also no differences in the proportions of each size class of lobsters measured at casitas with and without the caged Red Grouper after 48 hours (juveniles: T5 = -0.103, p = 0.920; sub-adults: T5 =

0.945, p = 0.367; adults: T5 = -1.19, p = 0.287; Figure 3.10). No lobster predators were observed

in or around any of the casitas during the experimental trials. In 2013 one of the casitas was occupied by an octopus, but this was discovered prior to starting any trials and so this casita was dropped from the study. All Red Grouper survived the 48-hour trials and were released to empty solution holes unharmed.

3.3.4. Estimated Lobster Survival Adjacent to Solution Holes

Out of the 60 lobster survival trials conducted in 2013, 38 of the lobsters (63%) were still tethered and alive after 24 hours. Four of the 60 trials were considered failures and were not included in survival probability calculations. Three of these trials were considered failed because the entire tether and loop were intact. Because of the way the tether was constructed, lobster escape due to experimenter error was determined to be the most likely cause of this outcome; however, because predation cannot be ruled out in these instances, these four trials were dropped

76 from the analysis. The fourth failed trial was excluded because the entire tether, lobster, and anchor was removed by an unknown party, most likely a local fisher who was observed diving in the area during the time the experiment was started. The remaining 56 trials ended after 24 hours with one of three outcomes: the lobster was present, alive, and unharmed (38 of 56 trials); the lobster was missing and the tether was snapped or the swivel was bent open (a predation event:

16 of 56 trials); or the lobster was absent, the tether was intact, and part of the carapace still attached to the tether loop (a triggerfish predation event: 2 of 56 trials). Triggerfish consume lobsters in a very specific way by first attacking the eyes and antennae (Bouma 2006). In both trials where triggerfish predation was suspected, the entire dorsal carapace, minus eye-stalks and antennae, was still attached to the tether after 12 hours, suggesting a crepuscular triggerfish predation event (see Appendix A). Furthermore, in both cases an adult triggerfish was observed at the solution hole at the 12-hour check, suggesting the lobster was consumed by the triggerfish rather than another predator.

2 After 24 hours, mean survival differed significantly across the three size classes (χ (2) =

17.5; p = 0.000157; Figure 3.11). Just 28.6% of the tethered juvenile lobsters were still alive after 24 hours, compared to 62.5% of sub-adults and 95% of adult lobsters. There were no

2 differences between 12 and 24 hour survival for any size class (juvenile: χ (1) = 0.0334; p =

2 2 0.855; sub-adult: χ (1) = 0.315; p = 0.575; adult: χ (1) = 0.00; p = 1.00). A survival curve was generated for tethered lobsters based on survival results. The estimates for the β0 and β1 parameters were -4.19 and 0.0846, respectively. Using the fitted logistic curve, estimated 50% survival of lobsters occurred at 4.96-cm CL (Figure 3.12).

77 3.4 Discussion

Predation is one of many factors that can shape the distribution of a prey species across a

landscape. Here I show how the presence of a predator, Red Grouper, had a significant effect on

the size distribution of its prey, lobsters, where they co-occur in solution holes of Florida Bay.

The overall effect of Red Grouper presence on total lobster abundance in solution holes was positive, but the specific effects of Red Grouper presence on lobsters varied across the different ontogenetic stages of lobsters investigated in this study. While both adult and sub-adult lobsters were more abundant in solution holes with Red Grouper, juvenile lobsters were less abundant when Red Grouper were present, and increased in abundance when Red Grouper were excluded from solution holes. This result generally aligns with the predictions for a predator-prey species interaction, where prey abundance should increase in the absence of the predator. The results of both the tethering experiment and the casita recruitment experiment support the hypothesis that the negative interaction between Red Grouper and juvenile lobsters is driven by predation.

Tethering results indicated that juvenile lobsters near Red Grouper occupied solution holes were unlikely to survive 24 hours, and despite evidence that lobsters actively avoid dens occupied by other piscine predators, the results of the casita experiment described here suggest that juvenile lobsters do not avoid dens occupied by Red Grouper.

The abundance of both sub-adult and adult lobsters declined in solution holes where Red

Grouper were excluded. In addition to their role as predators, Red Grouper also manipulate habitat by excavating sediment and detritus from Florida Bay solution holes (Coleman et al.

2010). Over time this excavating activity increases the depth of solution holes over time, and prevents them from becoming filled in; deeper solution holes hosted significantly more individuals, a pattern that was driven largely by the increase in lobsters (see Chapter 2), and that

78 was corroborated here in the significant relationships found between solution-hole depth and

lobster abundance. A number of empirical studies have found that chemical odors emitted by

lobsters attract conspecifics (Childress and Herrnkind 1996; Butler et al. 1999; Nevitt et al.

2000), and that the strength of attractant effects increase with increasing lobster density

(Eggleston and Lipcius 1992; Ratchford and Eggleston, 1998). By excavating solution holes,

Red Grouper increase the amount of habitat available to lobsters. Deeper solution holes that host

more lobsters produce more powerful attractant cues, which may explain the increased lobster

abundance in Red Grouper occupied solution holes compared to holes without Red Grouper.

Red Grouper were observed removing lobster molts from solution holes on multiple

occasions, and the presence of lobster molts outside of solution holes was a good indicator of the

presence of Red Grouper. While significant work has focused on the various cues and cycles that regulate metamorphosis and ecdysis, or molting, in crustaceans (see reviews by Charmantier et al. 1991; Lachaise et al. 1993; Chang 1995), surprisingly little work has been done on the fate of crustacean molts. In other crustacean species, shed molts are consumed by the molted individual

(Greenaway 2003). Crustacean exoskeletons contain large amounts of calcium and phosphate salts which can be reused in the formation of the new exoskeleton following ecdysis (Welinder

1974). Red Grouper were often observed removing lobster molts from solution holes, which could be detrimental to lobsters if they are consumed post-ecdysis to replace lost nutrients.

Conversely, removal of lobster molts from solution holes by Red Grouper could represent an attractant cue for lobsters searching for diurnal habitats; however the attractant potential of shed exoskeletons on lobster den selection has not been tested.

Because predator avoidance by juvenile lobsters has previously been experimentally verified for some lobster predators (e.g. nurse sharks [Eggleston and Lipcius 1992]; octopus

79 [Berger and Butler 2001; Butler and Lear 2009]), I felt it was necessary to directly test avoidance

by juvenile lobsters to Red Grouper. If juvenile lobsters avoid solution holes with Red Grouper,

this would serve as a valid alternate explanation for the negative interaction measured by the

exclusion experiment. The casitas that were constructed for this study were designed to provide shelter for juvenile and small sub-adult lobsters and to exclude both large lobsters and large lobster predators. Previous studies that tested the scaling of artificial shelters using casitas of similar construction found that limiting the size of den openings to less than 6-cm resulted in recruitment of primarily juvenile and small sub-adult lobsters (Eggleston et al. 1990; Eggleston and Lipcius 1992; Mintz et al. 1994; Arce et al. 1997). The addition of the empty fish cage to the

“control” casita was done to ensure that the quality of both experimental casitas was the same, so the only difference between casitas was the presence of the Red Grouper. The results of the six trials conducted in 2012 and 2013 provided no evidence to suggest that juvenile lobsters avoid dens with Red Grouper. Further study into this effect, for example a Y-maze choice experiment like the one used by Ratchford and Eggleston (2000) where lobsters choose sides of a tank fed with water from one of two head-tank sources, could be beneficial in verifying the lack of avoidance by juvenile lobsters towards Red Grouper.

The tethering experiment was conducted for multiple purposes: to directly estimate size- class-specific survival of lobsters next to solution holes, and to estimate a survival curve which may reveal a potential size refuge from Red Grouper predation. Similar tethering experiments have previously been used to measure the relative survival rates of lobsters and other shellfish

(e.g. Lipcius et al. 1998; Hovel and Fonseca 2005; Selgrath et al. 2007), but this methodology likely underestimates actual survival rates because the tether prevents escape from predators.

Because of this, tethering experiments should be limited to comparing relative mortality risk, or

80 “predation potential” between groups, and not for estimating absolute mortality (Peterson and

Black 1994; Aronson and Heck 1995; Selgrath et al. 2007). The size-class-specific survival estimates from the tethering experiment were consistent with the interaction strength estimates for each stage, with juvenile lobsters surviving least often, followed by sub-adults, and adult lobsters. While almost all of the adult lobsters survived the tethering trials (only one of 20 did not survive 24 hours), it was rare for a juvenile lobster to survive for 24 hours. Studies of

American lobster (Homerus americanus) have shown that larger lobsters enter a size refuge from predators that corresponds to increasing ranging behavior (Wahle 1992). A similar situation may exist in this system, where larger sub-adult and adult lobsters inhabiting solution holes experience lower predation risk than smaller conspecifics. The inflection point estimated by the survival curve from this study (4.96-cm CL; see Figure 3.12) roughly corresponds to the size at which sub-adult lobsters begin to forage more widely (4.5-cm CL). The size at which individuals enter a size refuge can have population-level consequences for lobsters, and will be a function of both lobster growth and predator size (Wahle 2003).

The size and abundance of lobsters in Florida Bay and the Florida Keys is regulated by natural predators, commercial and recreational fisheries harvest, and stochastic environmental events (e.g. harmful algal blooms, Butler et al. 1997). Several studies have shown that small spatial closures along the Florida Keys reef tract where lobster harvest is off limits, have resulted in larger and more abundant lobsters compared to nearby areas where harvest was allowed

(Bertelsen and Cox 2001; Cox and Hunt 2005). Unlike size and abundance, demographic transition rates are more likely to be regulated by within- and between-stage competition for resources (sensu de Roos and Perrson 2002). Maxwell et al. (2009) found that lobsters caught within the Dry Tortugas National Park (DTNP) had was significantly faster growth compared to

81 lobsters caught along the Florida Keys reef tract and in Florida Bay. The DTNP is located

approximately 100-km west of the Florida Keys and due in part to its isolation, experiences less

fishing pressure from commercial and recreational fishers compared to the Keys reefs and

Florida Bay. Diver surveys of fish populations along the Florida Keys reef tract have found that

the density of lobster predators, including Red Grouper, is larger in the DTNP compared to the

Florida Keys (Smith et al. 2011), and that differences in the density of lobster predators are even

greater within areas closed to fishing in the DTNP (Ault et al. 2006; Ault et al. 2013). Thus,

differences in the size-at-age of lobster populations that are separated spatially by only ~ 100-km

and that share similar abiotic conditions, may be explained by variation in ontogenetic-stage

transition rates resulting from differences in size-selective predation between populations (De

Roos and Perrson 2002; Miller and Rudolf 2011).

Between-stage competition for resources can ultimately reduce demographic transition

rates and lead to overall smaller populations. In one such example, Samhouri et al. (2009)

modified competition for shelter between adult and juvenile gobies by experimentally

manipulating the density of predator refuges. They found that by increasing the number of

predator refuges, thus reducing between-stage competition, this reduced the intensity of density-

dependent mortality in the smaller juvenile gobies (Samhouri et al. 2009). The interaction

considered in the present study, between Red Grouper and lobsters, highlights the complexity of effects that a predatory engineer can have on an ecosystem. Red Grouper excavation activity that increases the amount of lobster habitat may, as in the case with gobies, reduce competition for space and increase demographic rates allowing lobsters to more quickly enter a size refuge.

However, I also found that there was a positive relationship between the size of Red

Grouper and lobsters in solution holes (see Figure 3.8). The size at which lobsters enter a size

82 refuge from Red Grouper predation likely depends on the size of the predator. Red Grouper in

Florida Bay are also the target of an intense recreational fishery, and it is rare to encounter an individual larger than the minimum size limit of 20” TL (50.8 cm). The largest Red Grouper that

I encountered during this study was 64-cm TL, but the mean size was just 48.6-cm TL. By effectively limiting the size of the predator in a size-selective predator-prey interaction by

selecting for smaller predators, recreational fishers indirectly alter this interaction. This will effectively reduce the size at which lobsters must grow to enter the size refuge and may result in higher between-stage competition and ultimately slower lobster growth rates. Modeling studies aimed at investigating the population-level effects of these various interactions could be highly beneficial towards a better understanding of the indirect and multi-species effects of fishing in this ecosystem.

Size-selective predation can result in complex trophic interactions, which is especially evident in this case between Red Grouper and Caribbean Spiny Lobsters. As lobsters grow between ontogenetic stages, their interaction with Red Grouper shifts from a predator-prey interaction to a potentially mutualistic interaction. The size at which this interaction changes can have important population-level effects, which may even extend to the commercial and recreational fisheries that harvest both lobsters and Red Grouper from Florida Bay.

83 Table 3.1. Model selection results and description of the final “best-fit” models for linear mixed- effects model analysis of total lobster abundance and mean lobster size (CL) from lobsters counted and measured during diver surveys of Florida Bay solution holes. The “full” model included all listed factors plus solution-hole identity which was included in all models as a random variable. Results of the LRTs indicate the effect of either adding the interaction term or dropping each variable in question from the full model. Significant results indicate better model fit based on Chi-square analysis of the likelihood ratio statistics. Factors that significantly improved the model fit are shown in bold; if including the interaction term improved the model, then grouper presence (“grouper”) and “site” were not tested individually. AIC – Akaike’s Information Criterion – values are shown for each model to verify that the best fitting models had the lowest AIC values.

Dependent Variable Model Chi-sq df p AIC Total Abundance Full 369.15 + grouper * site 1.83 2 0.40 371.32 Final model: grouper 4.55 1 0.032 371.70 Abundance = site 12.6 2 < 0.01 377.73 grouper presence + site + year 1.00 1 0.61 366.15 area + depth + (hole) area 23.4 1 < 0.01 390.59 depth 120 1 < 0.001 484.33 Final 366.15

Mean Size Full 145.41 + grouper * site 3.64 2 0.16 145.78 Final model: treat 0.52 1 0.47 143.93 Mean size = site 1.54 2 0.46 142.95 area + depth + (hole) year 0.094 1 0.76 143.51 area 13.7 1 < 0.01 157.17 depth 9.43 1 < 0.01 152.84 Final 139.33

84

Figure 3.1. Lobster abundance at solution holes with (N = 53) and without (N = 46) Red Grouper (A.) and at three different sites (NBP = 8; NHC = 6; NSM = 13) in Florida Bay (B.) between 2010 and 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles. P-values shown are the results Chi-square likelihood ratio tests comparing a linear mixed model including the factor shown to a model without the factor shown.

85

Figure 3.2. Lobster abundance by solution-hole area (A.) and maximum excavated depth (B.), and mean lobster size by solution-area (C.) and maximum excavated depth (D.). Lobsters were counted and measured in situ at solution holes in Florida Bay between 2010 and 2013. Red Grouper presence at the time of the survey is indicated by filled circles. Regression lines shown are results of simple linear regressions and indicate significantly positive relationships for all four metrics analyzed: abundance = 6.36 + 1.40 * area; abundance = 1.65 + 0.23 * depth; mean size = 4.05 + 0.377 * area; mean size = 2.78 + 0.0601 * depth.

86 Table 3.2. Model selection results and description of the final “best-fit” models for linear mixed- effects modeling for the abundance of each of the three lobster size classes counted during diver surveys of Florida Bay solution holes. The “full” model included all listed factors plus solution- hole identity which was included in all models as a random variable. Results of the LRTs indicate the effect of either adding the interaction term or dropping each variable in question from the full model. Significant results indicate better model fit based on Chi-square analysis of the likelihood ratio statistics. Factors that significantly improved the model fit are shown in bold; if including the interaction term improved the model, then grouper presence (“grouper”) and “site” were not tested individually. AIC – Akaike’s Information Criterion – values are shown for each model to verify that the best fitting models had the lowest AIC values.

Juvenile Abundance Full 162.34 + grouper * site 8.43 2 0.014 157.91 Final model: grouper ------Juvenile abundance = site ------grouper * site + area + year 3.31 1 0.069 159.22 depth + (hole) area 11.4 2 0.003 165.33 depth 6.80 2 0.033 160.70 Final 159.22

Sub-adult Abundance Full 203.41 + grouper * site 4.21 2 0.12 203.20 Final model: grouper 8.60 1 0.003 210.01 Sub-adult abundance = site 2.90 2 0.23 202.31 grouper + area + year 0.96 1 0.33 202.37 depth + (hole) area 15.0 2 < 0.001 214.39 depth 6.59 2 0.037 206.00 Final 201.34

Adult Abundance Full 159.41 + grouper * site 0.37 2 0.83 163.05 Final model: grouper 3.62 1 0.057 161.04 Adult abundance = site 0.014 2 0.99 155.43 grouper + area + year 1.57 1 0.21 158.99 depth + (hole) area 10.7 1 0.001 168.10 depth 4.89 1 0.027 162.31 Final 155.00

87

Figure 3.3. Lobster abundance by size class and total lobster abundance with (N = 53; grey boxes) and without (N = 46; white boxes) Red Grouper. Lobsters were counted and measured in situ at solution holes in Florida Bay in 2012 and 2013. The dimensions of each box are delineated by the first and third quartiles, the line inside each box indicates the median value, the extended bars indicate the 5th and 95th percentiles, and the open circles indicate outliers beyond those percentiles.

88 55 With 50 W/O

45

40

35 Lobster abundnace 30

25 0 1 2 3 4 Week

Figure 3.4. Mean lobster abundance in solution holes with Red Grouper (filled circles; n = 15) and in solution holes where Red Grouper were excluded (open circles; n = 15). Data from experiments conducted in Florida Bay in 2011 and 2012. Error bars ± 1 S.E.

7.3 With W/O 7.1

6.9

6.7 Mean lobster CL [cm] CL lobster Mean

6.5 0 1 2 3 4 Week

Figure 3.5. Mean lobster size measured as carapace length (CL) in centimeters over time in solution holes with Red Grouper (filled circles; n = 15) and solution holes where Red Grouper were excluded (open circles; n = 15). Data from experiments conducted in Florida Bay in 2011 and 2012. Error bars ± 1 S.E.

89 1.0

0.8

0.6 Adult Sub 0.4 Juv

0.2 Proportionlobster classof size 0.0 Control Removal

Figure. 3.6. Proportion of lobsters by size class in solution holes with Red Grouper (“Control”; N = 15) and solution holes where Red Grouper were excluded (“Removal”; N = 15) as measured in Florida Bay in 2011 and 2012.

10 5

8 4

3 6 PI

PI 2 4

1 2

0 0

-1 -2 Juvenile Sub-adult Adult Total

Figure 3.7. Mean Paine’s Index values for lobster size classes and all lobsters (“Total”), calculated as the average change in abundance of lobsters in solution holes with (n = 15) and without (n = 15) Red Grouper in 2011 and 2012. Points are bootstrapped means; error bars represent 95% confidence intervals based on bootstrap results. Axes are scaled to better show error bars.

90 9.0

8.0

7.0

6.0 Mean Mean lobster(cm) CL 5.0

4.0 30 35 40 45 50 55 60 65 Red Grouper TL (cm)

Figure 3.8. Relationship between Red Grouper size and the mean size of all Caribbean Spiny Lobsters in solution holes when Red Grouper were removed in 2011 (circles; n = 9), 2012 (diamonds; n = 9), and 2013 (triangles; n = 8). All three years were pooled for analysis of the linear trend: mean lobster size = 0.0367 (Red Grouper size) + 5.3441.

18 60 A. p = 0.285 B. p = 0.938 15 55 12

9 50

6

45 Lobster(cm) CL

Lobster abundance 3

0 40 With Without With Without

Caged Red Grouper Treatment

Figure 3.9. Abundance (A.) and mean size (B.) of lobsters collected from casitas after 48 hours with (filled points) and without (open points) a Red Grouper inside a cage attached to the top of the casita. Results shown are means ± SE from six independent trials conducted in 2012 and 2013. P-values shown are results from two-tailed t-tests.

91 1.0

0.8

0.6 Adult Sub-adult 0.4 Juvenile

0.2 Proportionlobster classof size 0.0 Without With Caged Red Grouper Treatment

Figure 3.10. Proportion of lobsters by size class at casitas with and without a Red Grouper in a cage attached to the top of the casita after 48 hours. Proportions are pooled results from 6 independent trials conducted in 2012 (n = 2) and 2013 (n = 4).

100 12 hours 80 24 hours

60

* 40 % survival% *

20

0 Juvenile Sub-adult Adult Lobster size class

Figure 3.11. Survival of lobsters tethered adjacent to solution holes occupied by Red Grouper after 12 and 24 hours. Stars indicate groups that were significantly different (p < 0.05) from other size classes in same time period according to pairwise χ2 test with Bonferroni correction.

92 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 Surivialprobabilty 0.2 0.1 0.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 Lobster size (cm CL)

Figure 3.12. Survival curve for lobsters tethered adjacent to Red Grouper occupied solution holes in Florida Bay in 2013. Points represent binary results of each tethering trial (n = 56). The curve shown is the best-fit logistical regression of the tethering results. The dashed line represents the 50% survival probability at 4.96-cm CL.

93 CHAPTER FOUR

USING INVASIVE SPECIES TO TEST A BEHAVIORALLY-MEDIATED INDIRECT INTERACTION IN FLORIDA BAY

4.1 Introduction

Organisms interact in a variety of ways and the sum of these interactions form complex webs that constitute communities. Interactions between species are classified as direct, when one species affects the abundance or survival of another, or indirect, when the effect of a species on another is mediated through a third species (Abrams 1995). Indirect interactions can be further classified by the type of change that occurs in the intermediate species, either through changes in the abundance, or through changes in the traits or behaviors of the intermediate species (Strauss

1991). Trophically-linked indirect interactions, called “interaction chains” (Wootton 1993) or, more commonly, density-mediated indirect interactions (DMIIs; Abrams et al. 1996), include well-known examples like the keystone predators (e.g., seastars Pisaster ochraceus, and sea otters Enhydra lutris) that indirectly affect the abundance of resource species by consuming intermediate consumers (Paine 1969; Estes et al. 1978).

Indirect interactions mediated via changes in the traits or behaviors of the intermediate species are commonly called behaviorally-mediated indirect interactions (BMII; Miller and

Kerfoot 1987). BMIIs occur when the altered trait in question is behavioral, and often result from anti-predator behaviors that result in reduced foraging efficiency in the presence of a predator

(Abrams 1995; Lima 1998). A recent and well publicized example of a BMII followed the

reintroduction of wolves in Yellowstone National Park in 1995 (Ripple and Beschta 2012).

While the wolves did consume a small number of elk, their primary effect was to alter the

grazing patterns of the large ungulates, resulting in increased recruitment and survival of young

94 aspen trees inside the park. Similar trends have been reported following wolf reintroductions in

other Canadian and U.S. national parks (White et al. 1998; Beschta and Ripple 2009).

Studies that have compared the relative strengths of direct and indirect interactions have generally found that indirect interactions are often as significant as direct interactions for structuring communities (Miller 1994; Menge 1995; Wilbur 1997; Werner and Peacor 2003). In some cases, the magnitude of indirect interactions can even negate the effect of direct interactions, a pattern that is true for indirect effects that occur both within and between trophic levels (Miller 1994; Wootton 1994). Indirect interactions can even drive trophic cascades when prey species alter behaviors enough to cause changes in the density of resources (Schmitz et al.

2004). Given the complexity and importance of indirect interactions in structuring communities, a complete understanding of ecological communities can only be developed by investigating both the direct and indirect interactions that occur within communities.

The fundamental difference between DMIIs and BMIIs is the type of effect initiated on the intermediate species. In DMIIs, these effects are generally driven by predator-prey or competitive interactions, on inter- and intra-trophic level dynamics, respectively. BMIIs occur when visual, chemical, or tactile cues of predators affect the behaviors of prey. For example, oyster survival on intertidal oyster reefs is enhanced by the presence of piscine predators that suppress foraging activity by mud crabs (Grabowski 2004; Hughes et al. 2012). When fish predators were removed, the mud crabs consumed more oysters causing lower oyster survival, except where mud crab density was reduced by the presence of blue crabs, an alternate predator

(Grabowski et al. 2008). Thus, enhanced oyster survival in the presence of toadfish or catfish is the result of BMIIs, whereas enhanced oyster survival in the presence of blue crabs is the result of a DMII. Considering the role of BMIIs in predator-prey dynamics can reveal important ways

95 in which predators can structure ecosystems without directly consuming their prey (Peckarsky et

al. 2008).

A number of classic studies in ecology have shown the importance of predators in

shaping the structure of prey communities across diverse aquatic and marine ecosystems

including in ponds (Brooks and Dodson 1965), along the rocky intertidal (Paine 1966; Menge

1976), and on coral reefs (Hixon 1991). More specifically, substantial evidence supports the ability of resident fish predators to affect the size and structure of reef fish communities via direct piscivory on post-settlement fish recruits (Hixon and Beets 1993; White 2007; Stallings

2009; Stier et al. 2013). Studies of BMIIs caused by fish predators tend to focus on the indirect

effects of fishing on marine fish communities (e.g. Dayton et al. 1995; Jennings and Kaiser 1998;

Dulvy et al. 2004).

Among the few studies to explicitly test piscivory-driven BMIIs is Stalling’s (2008) work

in the Bahamas in which he found a strong, positive BMII driven by changes in the foraging

behavior of intermediate predators (small-bodied groupers) in the presence of the larger Nassau

Grouper (Epinephelus striatus). Pusack (2013) investigated a similar interaction in which he

found evidence for a BMII between Nassau Grouper and juvenile coral reef fishes mediated by

the behaviors of lionfish (Pterois volitans and P. miles), a mesopredator native to the Indo-

Pacific and invasive in the western Atlantic since at least 1985. Lionfish have experienced a

rapid range expansion throughout the Caribbean, Gulf of Mexico, and southeastern US Atlantic

coast since about 2004 (Coté et al. 2013). They are generalist predators that consume a diverse

array of fishes (Morris and Akins 2009; Munoz et al. 2011; Valdez-Moreno et al. 2012; Coté et

al. 2013), and are able to reduce the abundance of native western Atlantic coral reef fishes by 80

to 94% and the biomass by up to 65% (Albins and Hixon 2008; Green et al. 2012; Albins 2013).

96 The invasion of lionfish in the Florida Keys is more recent relative to their spread through the Caribbean. The frequency of diver-reported encounters along the Florida Keys reef tract increased 250-fold between 2009 and 2011 (Schofield 2009; Ruttenberg et al. 2012). By

2010, they had invaded the hardbottom habitats of Florida Bay, which is adjacent to the Keys

(Figure 2.1). While trophic studies of lionfish diet studies indicate that they consume fish species associated with hardbottom habitats in Florida Bay (Layman and Allgeier 2012; Coté et al.

2013), no studies have yet investigated the effects of lionfish on fish communities in this habitat.

Florida Bay serves as an important nursery habitat for many fishes and invertebrates, which, as they mature, move to nearby coral reefs or hardbottom habitats further offshore as adults

(Fourqurean and Robblee 1999). The lionfish invasion of Florida Bay could have significantly repercussions on local reef fish populations.

One of the mesopredators that interacts with lionfish in Florida Bay is the Red Grouper.

In Florida Bay, Red Grouper are primarily associated with karst hardbottom features called solution holes – areas where freshwater has dissolved away the limestone leaving behind pits or holes in the bottom. These holes are actively excavated by Red Grouper and subsequently colonized by a suite of fish and invertebrate species (Coleman et al. 2010). Previous experiments conducted in Florida Bay solution holes showed that Red Grouper presence in solution holes positively affected the abundance and diversity of associated faunal communities (see Chapter

2). However, these community-level effects were driven by strong interactions with only a small

number of individual species from the total species pool. This group included juvenile coral-reef

fishes, primarily small juvenile grunts (Haemulon spp.), which were consistently the most

numerous fauna encountered in solution holes. Small juvenile grunts are primarily planktivorous

97 (Cocheret de la Morinière et al. 2003) and juvenile grunts encountered in Florida Bay were most

often found hovering in the water column above hardbottom habitat features.

Red Grouper primarily consume crustaceans and demersal fishes (Randall 1967; Moe

1969; Bullock and Smith 1991; Brule et al 1993; Weaver 1996). They do not feed on juvenile

grunts or any of the locally abundant piscivore species, and do not appear to have any effect on the abundance of resident piscivores in solution holes (see Chapter 2). Thus, given no clear

density-driven mechanism to explain the strong positive interaction between Red Grouper and

juvenile grunts found in Chapter 2, I hypothesize that this pattern results from BMIIs of Red

Grouper via solution-hole associated piscivores.

Red Grouper are almost always the largest animal encountered in solution holes, and as such, could displace or disrupt piscivory by resident and transient piscivores around solution holes. Over time, solution holes with Red Grouper result in the differential survival of post- settlement juvenile reef fish when compared to their survival in holes without Red Grouper. To test this hypothesis, I used the presence of invasive lionfishes in Florida Bay as a proxy for native piscivores. While other large-bodied groupers, including Nassau and Tiger Grouper

(Mycteroperca tigris), will prey on lionfish (Maljkovic et al. 2008), Red Grouper do not

(Coleman, pers. comm.; pers. obs.). Furthermore, Nassau and Red Groupers are extremely similar in appearance, and given that evidence for changes in lionfish behavior in the presence of

Nassau Grouper has been reported elsewhere (Pusack 2013), I felt it reasonable to hypothesize that lionfish will alter their behavior in the presence of Red Grouper, even if they do not, in fact, present much actual predation risk.

Using lionfish to test this potential BMII had several distinct advantages over manipulating native piscivore density. First, compared to the abundance of piscivores associated

98 with solution holes, the abundance of lionfish in Florida Bay was much greater. Most of the

piscivores in Florida Bay (e.g. jacks, barracudas, etc.), are roving or transient, rather than

associated with specific solution holes, although they routinely passed in the vicinity of holes.

Second, lionfish prefer hardbottom and reef habitat (Biggs and Olden 2011), exhibit high site

fidelity (Jud and Layman 2012), and appear to gravitate to solution holes (Coleman, pers. comm.; pers. obs.), making them relatively easy to locate and capture. Because solution holes are effectively isolated patches which, in Florida Bay are surrounded by seagrass, it seemed realistic

to assume that lionfish transplanted to solution holes would remain there. Third, lionfish invaders

have large negative effects on native fish populations relative to native piscivores. For example,

Albins (2013) found that lionfish predation on Bahamian coral reef fish populations were 2-3

times that of native piscivores. Thus, if the effects of lionfish in Florida Bay are consistent with

those found elsewhere, then this should increase the ability to detect differences in juvenile reef

fish abundance that are attributable to the lionfish versus native piscivores, and make any BMII

effects easier to detect.

Finally, because disentangling the relative strength of DMIIs and BMIIs is not a trivial

task, using lionfish as a proxy for native piscivores offers a distinct advantage in terms of the

experimental design. Experiments designed to determine the strength of indirect interactions

between a predator and a prey should include four treatments: (1) a predator-present treatment,

where the predator and prey are both present; (2) a predator-absent treatment, where the prey is

present without the predator; (3) a threat treatment, where predator cues are present but direct

predation of prey is prevented; and (4) a culling treatment, where prey are experimentally

removed to simulate predation in the absence of predator cues, analogous to a control where

neither prey nor predators are present (Okuyama and Bolker 2007). This experimental design

99 ensures that DMII and BMIIs are estimated together (with the true predator treatment), in the

absence of indirect interactions (with the predator-absent treatment), and separately (with the

culling and threat treatments, respectively), thus enabling quantification of both types of indirect

interactions.

I assumed that Red Grouper were unlikely to prey on lionfish, but that their presence provided the threat of predation, making the true-predator and threat treatments functionally

equivalent. This assumption allowed me to ignore any DMIIs and just test for BMIIs between

Red Grouper and the suite of solution-hole associated juvenile coral reef fishes via lionfish. To accomplish this, I conducted the following experiments: (1) Red Grouper present-absent

experiments to quantify the effects of Red Grouper on solution-hole associated juvenile reef fish

abundance and diversity; (2) lionfish present-absent experiments to quantify the effects of

lionfish on solution-hole associated juvenile reef fish abundance and diversity; and (3) lionfish

present with and without Red Grouper to quantify the effects of lionfish on juvenile reef fish

recruitment in the presence and absence of Red Grouper.

4.2 Methods

4.2.1. Study Site

Florida Bay is a large open embayment in south Florida bordered by the Florida Keys, the

Everglades, and the Gulf of Mexico (Figure 2.1). The benthic community of Florida Bay is

dominated by seagrasses, which covers ~75% of the bay bottom (Fourqurean et al. 2002).

Interspersed between seagrass beds are areas of karst hardbottom – exposed limestone

formations that are usually covered in a thin veneer of sediment and pockmarked with solution

holes. For this study, I used a set of 31 solution holes located at the Wilkerson South site located

100 in outer Florida Bay (Figure 2.1). I selected a subset of solution holes (N = 18) that appeared

similar in terms of size and location to minimize variation in solution-hole features (Table 4.1).

Solution hole area, defined as the product of the two longest perpendicular measurements,

ranged from 1.69-m2 to 6.99-m2 (mean ± SEM = 4.44 ± 0.372 m2); maximum excavated depth, defined as the deepest single measurement taken within the excavated area of the solution hole ranged from 26-cm to 77-cm (45.8 ± 2.70 cm).

4.2.2. Red Grouper Effect Experiment

I ran a 4-week field experiment to test the effects of Red Grouper on the abundance and

diversity of juvenile reef fish associated with solution holes. I first surveyed of all 15 solution

holes to assess Red Grouper presence. Fourteen of the 18 holes were occupied by a Red Grouper

in 2012. Of these, I randomly selected 10 Red Grouper occupied solution holes and assigned

them to one of two treatment groups: Red Grouper present (n = 5; RG+), or Red Grouper absent

(n = 5; RG-), the latter of which would have Red Grouper removed using hook-and-line,

measured to the nearest cm TL, tagged with a dart tag, and released a vacant solution hole

located at least five-km from the experimental study area. Translocated fish were not observed

again at experimental holes during the experiment. I visited each of the five RG- solution holes

at least once every 48 hours for the duration of the experiment to check for the presence of a new

Red Grouper. If experimental holes were recolonized by a new Red Grouper, the fish was

captured, measured, tagged, and released at a new site as described above; this occurred 11 times

during the experiment.

Prior to removing any Red Grouper from the assigned RG- solution holes, a team of two

divers on SCUBA conducted a census of all reef fishes associated with all 10 solution holes,

following the methods described by Hixon and Beets (1993). After slowly approaching the

101 solution hole to a distance of ~1-m from the edge, each diver slowly circled the hole while

recording the number and identity of each fish species on an underwater slate. Divers first

focused on the active planktivores hovering above the hole, then enumerated and recorded the

demersal and cryptic species and macroinvertebrates inside the hole using flashlights to aid

identification. Recorded abundances were averaged for each species across divers, and total species richness was determined as the sum of all distinct species observed by both divers.

Divers visually estimated the size of each fish to the nearest 1-cm (below 10-cm TL) or to the

nearest 5-cm (above 10-cm TL). Each survey lasted until each diver had examined the entire

solution hole; the mean duration of diver censuses was about 12 minutes. Because the grunt

species complex in Florida Bay contains at least six different species that are visually

indistinguishable at sizes < 5-cm TL, grunts were identified to species when possible whereas all

individuals < 5-cm were grouped together as “grunt recruits”. This same protocol was repeated

weekly for 4 weeks (5 total surveys at each of the 10 holes). On average, a complete survey of all

ten holes took approximately 2 days.

4.2.3. Lionfish Effect Experiment

To test the effects of lionfish on the abundance and diversity of juvenile reef fish

associated with solution holes in the presence and absence of Red Grouper, I repeated the

experiment described above in 2013 with lionfish included as an additional factor. In June 2013,

the initial snorkel survey revealed that 14 of the 18 solution holes surveyed in 2012 were

occupied by Red Grouper, and three additional occupied solution holes were found (total

occupied holes in 2013 = 17). I randomly selected 16 Red Grouper occupied solution holes and

assigned them to one of four treatment groups (n = 4): (1) no predators (NP); (2) lionfish only

(LO); (3) Red Grouper only (RG); and (4) lionfish and Red Grouper (L+RG). Here, the term “no

102 predators” means no Red Grouper and no lionfish; other predators were not manipulated. In all

other relevant treatments, only a single individual lionfish or single Red Grouper was used, in

keeping with typical densities of these fish observed in Florida Bay. Red Grouper are generally

solitary and display territorial aggressive behaviors towards conspecifics. Lionfish densities

observed in Florida Bay from 2010 to 2012 were similarly low, and despite increasing in

occurrence over time, it was rare to encounter more than a single individual lionfish in a solution

hole. For the purposes of estimating the BMII between Red Grouper and juvenile reef fish via

lionfish, the NP treatment represents the baseline “control” condition, the RG and LO treatments

estimate the total effect of each predator separately on juvenile reef fish abundance, and the

L+RG treatment is the “threat” condition as described above.

At the start of the experiment there were no lionfish present at any of the 16 randomly- assigned experimental solution holes. I collected eight lionfish from hardbottom habitats located elsewhere in Florida Bay with monofilament hand nets, then immediately transported them to the study area, and released one at each of the appropriate assigned LO or L+RG solution holes. Red

Grouper occupying NP and LO solution holes were captured and relocated following the protocol described above in the Red Grouper effect experiment.

Following predator transplants, I visited NP and LO treatment holes once every 48 hours for the duration of the experiment to ensure that no other Red Groupers had moved into these sites. New individuals encountered were captured, measured, tagged, and released at unoccupied sites as described above; this occurred 10 times during the 2013 experiment. Weekly diver surveys were conducted at all 16 experimental solution holes for 6 weeks following the initial survey (7 total surveys per hole). At the end of the experiment all lionfish were collected with hand nets and euthanized with an overdose of MS222 (IACUC 2011). The average size of

103 lionfish used during the 2013 experiments was 18.6 ± 1.14-cm TL, which corresponds to an age

of approximately 5-6 months (Potts et al. 2011).

4.2.4. Statistical Analysis

I calculated a variety of community response variables to determine the effects of Red

Grouper and lionfish separately and in concert on the native juvenile reef fish populations

associated with solution holes. These were: total abundance (N), Hill’s diversity numbers H0, H1,

and H2, and Hill’s evenness (E). Hill’s numbers provide a means of calculating commonly used

diversity indices using the single equation:

1 S 1−a  a  H a = ∑ pi   i=1 

where pi is the relative proportion of the community made up by species i (Hill 1973). When

evaluated for integer values of a of 0, 1, and 2, Ha reduces to species richness, the antilog of the

Shannon-Weiner index, and the reciprocal of the Simpson’s Index, respectfully. Generally, as a

increases, the index gives greater weight to more abundant species. Hill’s evenness, E, was

calculated using the equation:

, = /

2 1 2 1 that Hill initially proposed because it does� not include� � species richness (H0) and therefore is

relatively insensitive to sample size. Hill’s evenness converges to 1 when all species are equally

abundant, so smaller values indicate more uneven communities.

I used linear mixed-effects models to draw inferences about the effects of predator

treatment on each of the five community metrics (N, H0, H1, H2, and E). The models included

two categorical explanatory variables, predator treatment (either [RG+ or RG-], or [NP, LO, RG,

or L+RG] depending on the year of the experiment) and week, plus a random intercept for each

104 solution hole. I included time (week) as a categorical variable to eliminate any assumptions about the relationship between response variables and time. Solution-hole identity was included as a random variable in all models due to the repeated measures design of the experiment that included multiple observations of the same hole across time. Model selection to determine if including a treatment by week interaction term improved the model was based on comparing likelihood-ratio tests (LRT) performed on nested models. If the interaction did not significantly improve the model, predator treatment and week were tested separately by dropping each term independent of the other. Visual examination of the residuals suggested violations of the assumptions of both homogeneity of variance and independence for at least some of the models, so I fit alternative models for each community response metric. These alternative models were: one that incorporated heteroscedasticity among treatments, one that incorporated temporal autocorrelation among observations within solution holes using the AR(1) autoregression model, and one that incorporated both variance and autocorrelation structures. The resulting models were compared using Akaike’s Information Criterion (AIC) to determine the optimal variance and autocorrelation structure. For both experiments I used pairwise t-tests to evaluate differences between treatments at the end of the experiment for each of the five community response metrics based on the optimal LMM models from the model selection. When testing for differences between treatments, I considered p-values less than 0.05 to represent strong evidence, whereas p- values between 0.05 and 0.10 were considered to represent marginal evidence against the null hypotheses.

To evaluate the effects of Red Grouper and lionfish on the structure of the juvenile reef fish communities, I performed ordinations with non-metric multidimensional scaling (NMDS).

For all ordinations, recruit abundances were first square-root transformed and then standardized

105 using the Wisconsin double standardization, where abundances were first standardized by

species maxima and then by the sample total. I used Bray-Curtis distance for both the ordinations

and for hypothesis testing of recruit community structure. To further detect differences in the

composition of recruit communities, I used permutational multivariate analysis of variance, or

PERMANOVA (Anderson 2001) to test for differences among treatments for both experiments.

Recruit abundances were not transformed or standardized, and all analyses were run with 1,000

unconstrained permutations.

Finally, I calculated effect sizes for the direct effects of lionfish, the total indirect effects

of Red Grouper, and the effects of both predators together on juvenile reef fish recruit abundance

using a ratio-based approach (Trussell et al. 2006; Hughes et al. 2012). The direct effect (DE) of lionfish on recruit abundance was calculated with the ratio of recruit abundance in the LO treatment to the mean recruit abundance in the NP treatment. Similarly, the indirect effect (IE) of

Red Grouper on recruit abundance was calculated with the ratio of recruit abundance in the RG treatment to the mean recruit abundance in the NP treatment; I calculated the IE of Red Grouper

on recruit abundance calculated separately for 2012 and 2013. Finally, the BMII of Red Grouper

on recruit abundance via lionfish was calculated with the ratio of the recruit abundance in the

L+RG treatment, the effect of both predators or the lionfish effect in the presence of predator

cues, to the mean recruit abundance in the LO treatment (following Okuyama and Bolker

[2007]). Thus for each case, in terms of recruit abundance, R, at the end of the six-week

experiment:

DELION = (RLO / RNP) – 1

IERG, YEAR = (RRG / RNP) – 1

BMII = (RL+RG / RLO) – 1

106 The numerators for all ratios were provided by all replicates of the given treatment, whereas the

denominator was the mean recruit abundance at the end of the experiment for the given

treatment. Means and 95% confidence intervals were estimated for all effect sizes by

bootstrapping 1000 times with replacement to account for the low number of experimental

replicates available for each of the treatments. This approach is similar to that used to quantify

interaction strength in Chapters 2 and 3, and by Paine (1992).

All statistical analyses were conducted in the R software environment (R Core Team 2014)

using the “lme4” package for the LMMs (Bates et al. 2012), the “MASS” and “vegan” packages for the NMDS and PERMANOVA analysis (Venables and Ripley 2002; Oksanen et al. 2011), and the “resample” package for the bootstrapping procedure.

4.3 Results

In both experiments, Red Grouper had a significant positive effect on the abundance of juvenile reef fish recruits. In 2012, Red Grouper presence resulted in 75.8 (± 28.3) more recruits compared to the holes where Red Grouper were excluded (t = -2.68; p = 0.0287; Figure 4.1). In

2013, after six weeks the abundance of juvenile reef fish recruits was greatest at the Red Grouper

holes (111 ± 35.4), followed by holes with both a lionfish and a Red Grouper (50.8 ± 15.2), and

holes with neither predator (43.8 ± 17.2). Of the four Lionfish Only holes, three had zero recruits

encountered after six weeks, while the fourth hole had 33 recruits present (mean LO recruit

abundance = 8.25 ± 8.25). In the 2013 experiment, the Red Grouper Only (RO) treatment again

resulted in significantly more recruits compared to the No Predator (NP) treatment, which was

fundamentally the same comparison as in the 2012 experiment (t = -1.94; p = 0.05; Figure 4.2).

Recruit abundance was also significantly greater at Red Grouper Only holes compared to

107 Lionfish Only (LO) holes (t = 3.26; p = 0.024; Table 4.2). When Red Grouper and lionfish were present together in solution holes, the abundance of juvenile reef fish recruits was significantly different from the Red Grouper only (t = 1.82; p = 0.059) and lionfish only treatments (t = -3.07; p = 0.027), but not significantly different from the No Predator treatment (t = 0.23; p = 0.41;

Table 4.2).

The results of the linear mixed models for the 2012 experiment showed that Red Grouper presence alone was a significant factor for explaining improved the model either alone, or as an interaction with time for recruit abundance, species richness, and evenness (Table 4.3). Allowing for different variance by treatment to account for heteroscedasticity improved only the model for species richness, while including temporal autocorrelation improved only the model for recruit abundance (Table 4.3). For the linear mixed model analysis of the 2013experiment, including treatment by week interaction improved the model fits for recruit abundance and evenness (Table

4.4). The preferred variance and autocorrelation structures varied among the metrics tested, where both were included in the optimal model for recruit abundance, only temporal autocorrelation was included in the optimal model for richness (H0), neither variance or autocorrelation structures were included in the optimal models for either diversity metric (H1 and

H2), and only variance by treatment included in the optimal model for evenness (Table 4.4).

Red Grouper and lionfish had varying effects on the other community metrics. In 2012,

excluding Red Grouper from solution holes resulted in 2.40 (± 0.704) fewer species of juvenile

reef fish compared to solution holes with a Red Grouper (t = -3.39; p = 0.011; Figure 4.3).

Recruit communities with Red Grouper were significantly less even compared to communities

without Red Grouper (t = 3.03; p = 0.019); however Red Grouper presence did not have a

significant effect on either the Shannon (t = -1.56; p = 0.160) or Simpson’s diversity (t = -0.948;

108 p = 0.373) of juvenile reef fish recruits. In 2013, Red Grouper alone did not have a significant

effect on any of the recruit community metrics tested (see Table 4.2). In addition to significantly reducing the abundance of recruits (see Figure 4.2), lionfish alone had a significantly negative effect on the species richness, Shannon and Simpson’s diversity, and evenness of the juvenile reef fish recruit community compared to the predator-free holes (Figure 4.4). Compared to

solution holes with Red Grouper alone, recruit communities were significantly less species rich (t

= 3.42; p = 0.014) and less even with lionfish alone (t = 3.00; p = 0.058). There was no

difference in either the Shannon (H1: t = 1.50; p = 0.208) or Simpson’s (H2: t = 1.50; p = 0.194) diversity measures between the Red Grouper only and lionfish only treatments (Figure 4.4).

In general, when Red Grouper and lionfish were both present in solution holes, the juvenile reef fish communities found there were more similar to holes without either predator than they were to holes with either a Red Grouper or lionfish alone. Compared to the No

Predator treatment, solution holes with both predators were not significantly different in terms of any of the community response metrics analyzed, including recruit abundance (see Figures 4.2 and 4.4). Compared to holes with a Red Grouper alone, recruit communities were significantly more species rich when both predators were present (5.75 [± 0.629] recruit species with both Red

Grouper and lionfish versus 3.25 [± 0.629] recruit species with only Red Grouper).

Recruit community structure was significantly different when Red Grouper were present.

During the course of the 4-week long experiment in 2012, 14 species of juvenile reef fishes were observed at experimental solution holes, although only 10 of these were identified during the final diver survey (Table 4.5). The NMDS ordination of recruit communities at the end of the

Red Grouper exclusion experiment showed a clear separation between solution holes with and without Red Grouper (Figure 4.6). This difference was supported by the results from the

109 PERMANOVA analysis which confirmed that Red Grouper presence had a significant effect on

the structure of solution-hole associated juvenile reef fish communities (pseudo-F1,9 = 3.40, p =

0.042). Differences in community structure were driven mostly by the grunts, specifically White

Grunts and unidentified grunt recruits (those < 5-cm TL; Table 4.3). Juvenile White Grunts were the most prevalent species in all communities, and were, on average, 156% more abundant in solution holes with Red Grouper. Only one species, Gray Angelfish (Pomacanthus arcuatus), was less abundant at the Red Grouper exclusion holes.

During the lionfish experiment in 2013, 14 species of juvenile reef fishes were encountered; however the suites of juvenile reef fishes were slightly different from those observed in 2012 (see Table 4.6). The NMDS ordination of recruit communities showed clear separation of communities with lionfish present (Figure 4.6). Because three of the four lionfish- only solution holes had no recruits at the end of the experiment, the three zero-abundance points

overlaid on top of each other (see point at [-1.33, 0.02] in Figure 4.6), and the line representing

the standard deviation of the lionfish-only communities collapsed to a line. There was significant

overlap in the Red Grouper only and no predator communities, while the communities with both

predators were separate from all other groups. The PERMANOVA results indicated that predator treatment had a significant effect on community structure (pseudo-F3,15 = 2.27, p = 0.029).

The estimated direct effect of lionfish on recruit abundance, calculated with the ratio of recruit abundance with lionfish to recruit abundance with no predators, was -0.802 [-1.00, -

0.434], indicating that the expected recruit abundance with lionfish alone was between zero and half that of the No Predator treatment (Figure 4.7). The estimated indirect effect (IE) of Red

Grouper on recruit abundance was 2.10 [0.966, 3.03] in 2012 and 1.546 [0.206, 2.77] in 2013, indicating that the expected recruit abundance in the presence of a Red Grouper was about twice

110 that of the No Predator treatment. The BMII between Red Grouper and juvenile reef fish recruits via lionfish was estimated as 5.18 [2.15, 8.73] or 5-times the expected recruit abundance with lionfish alone.

4.4 Discussion

The experiments conducted in 2012 and 2013 both showed that Red Grouper enhance the abundance of juvenile reef fishes that recruit to solution holes in Florida Bay, and supported the

hypothesis that this effect occurs via a BMII mediated through changes in piscivore behavior in

solution holes. Juvenile reef fishes benefited from the presence of the relatively large, territorial

habitat manipulator. Conversely, recruit abundance was significantly depleted with lionfish.

Lionfish are extremely efficient predators on Caribbean reef fishes, and coral reefs and

hardbottom habitats previously invaded by Lionfish have significantly depleted reef fish

populations shortly following the invasion (Albins and Hixon 2008; Munoz 2012). The results

presented here confirm that the negative effects of lionfish on native reef fish populations

observed on coral reefs and hardbottom habitats elsewhere in the invaded range also hold true for

Florida Bay hardbottom fish communities.

Lionfish and Red Grouper had very different effects on the diversity of juvenile reef

fishes in Florida Bay solution holes. In 2012, Red Grouper presence resulted in higher species

richness, and lower evenness of recruits, but did not significantly affect either Shannon or

Simpson’s diversity. In 2013 Red Grouper none of these effects were detectable when comparing

solution holes with and without Red Grouper. Meanwhile, similar comparisons are difficult to

make for lionfish as they reduced recruitment to zero at three of the four Lionfish Only treatment

holes. While all four community diversity metrics were depressed in the Lionfish Only treatment

111 compared to the others, it becomes difficult to make specific statements about changes to

diversity when the predator leaves no prey at all. Certainly this is a strong qualitative result, if

not a quantitative one: lionfish reduce the abundance and diversity of native Caribbean reef

fishes, apparently through indiscriminate piscivory.

The more interesting results from the 2013 experiment are effects of both Red Grouper and lionfish together. There were no differences in recruit abundance in solution holes with both predators compared to the other treatments, and solution holes with both predators had comparable species richness and diversity compared to the Red Grouper Only and No Predator treatments. These results support the BMII hypothesis: when Red Grouper were present in solution holes with Lionfish, the communities of juvenile reef fishes, although depressed in abundance, were not significantly different from those without Red Grouper or without either predator. Again, native fish communities benefited from the presence of Red Grouper, compared to when the exotic piscivorous lionfish was present alone.

Although Red Grouper do not compete with lionfish for prey, they may compete for habitat or be recognized by lionfish as a potential predator. The exact nature of the lionfish response to Red Grouper, either through some specific behavior modification by the lionfish, or a more generally disruptive effect of Red Grouper presence on piscivores, remains untested. Few other studies have investigated the behavioral interactions between Atlantic coral reef fishes and lionfish. A study of the response by three-spot damselfish, Stegastes planifrons, to invasive lionfish found a negligible response by the typically aggressive damselfish (Kindinger 2014).

However, it is important to note that damselfish used in Kindinger’s study were generally of equal or smaller size compared to the lionfish, while the Red Grouper used in the present study were 2-3 times larger in terms of total length. Another study that investigated competitive shelter

112 use between lionfish and Nassau Grouper in mesocosms, found that Nassau Grouper generally avoided lionfish, even when they were much larger than the lionfish, but that lionfish used shelters less frequently with Nassau of any size (Raymond et al. 2014). While the details of individual interactions between Red Grouper and lionfish were not rigorously detailed in this study, observations made by divers during fish censuses suggested that the lionfish may actively avoid Red Grouper. Divers observed that lionfish were most often encountered on the opposite side of the solution hole from the Red Grouper. However the duration of these observations were limited to the time spent conducting community censuses. Further investigation into avoidance behaviors exhibited by either fish may be easily quantified to determine the validity of the assumptions that lionfish actively avoid Red Groupers.

Recent experimental evidence suggests that larval reef fishes use complex chemical cues to guide settlement from planktonic to benthic stages (Dixson et al. 2012; Dixson et al. 2014).

Excavation activities by Red Grouper may contribute some unidentified chemosensory cues that lead to enhanced recruitment of larval reef fishes to solution holes when occupied by a Red

Grouper. Recruitment enhancement was not explicitly tested in these experiments, which were designed under the assumption that observed differences in the abundance of juvenile reef fishes was driven primarily by differences in piscivory. However, experimentally testing differential recruitment patterns in response to Red Grouper excavation activity may be a fruitful line of research. Most likely some combination of preferential recruitment and behavioral effects of Red

Grouper on resident and transient predators are responsible for the patterns shown here and elsewhere in this dissertation. The link between recruitment, post-settlement mortality, and adult population size for fish is complex; however, Shulman and Ogden (1987) found that changes in immediate post-settlement survival of French grunts was a more important factor regulating the

113 ultimate abundance of adult grunts on coral reefs compared to changes in recruitment. By

including lionfish in these experiments I was able to test the hypothesis that enhanced reef fish

recruitment at solution holes with Red Grouper was at least in part driven by behavioral

interactions leading to reduced piscivory.

I observed no predation on any of the transplanted lionfish during the experiment in

2013, despite the presence of native predators that have been cited as possible predators of

invasive lionfish (Mumby et al. 2012; Diller et al. 2014). All of the transplanted lionfish survived

the duration of the experiment and remained at the same solution hole where they were initially

transplanted. Despite increasing awareness of the effects of the lionfish invasion across the

invaded range, including attempts by spearfishers to “teach” native groupers and sharks to feed

on lionfish, ultimately it may be intact native fish communities that include native mesopredators

that compete with and alter the behavior of lionfish are the best chance of ameliorating the worst- case effects of lionfish (Hixon and Albins 2011). Some recent evidence suggests that Caribbean reefs with relatively high native predator density can maintain unchanged prey populations despite being invaded by lionfish (Elise et al. 2014). Other studies have found that that lionfish- induced reductions in the density of native fishes <10-cm TL did not translate to prey sized 10-

20-sm TL (Albins 2015). Lionfish, then, may represent an enhanced gauntlet that juvenile reef fish must pass through, but not necessarily an impenetrable one.

The invasion of lionfish in the Caribbean and western Atlantic has been swift and the effect on native reef fish communities has been severe (Cote et al. 2013; Albins and Hixon

2013). A possible mechanism that may explain the success of this invasion is the release from competition and predation that lionfish experience in the novel ecosystem, also known as meso- predator release (Mack et al. 2000; Brashares et al. 2010). Humans have reduced the densities of

114 native meso-predators on coral reefs and hardbottom habitats throughout the Caribbean, resulting in fewer competitive interactions that serve to limit both the population size and predatory effects of lionfish in its native range (Stallings 2009b; Albins and Hixon 2013). The experiments presented suggest that here support this meso-predator release hypothesis by showing how the presence of a native predator, the Red Grouper, can partially ameliorated the negative effects of lionfish predation on native reef fish communities.

Collecting information that quantifies the community-level effects of fisheries species is an integral part of expanding fisheries management from single-species stock assessments to ecosystem-based fisheries management (Pikitch et al. 2004). In this case, Red Grouper, an important fishery species in the southeast US, has complex indirect interactions with the other species that colonize grouper-excavated solution holes in Florida Bay and presumably across the rest of the Red Grouper’s range. Some of these interactions encompass multiple trophic levels and may have population-level effects on other species that support fisheries or provide important ecosystem services. Currently, the Gulf of Mexico Red Grouper population is declining, though technically neither overfished nor undergoing overfishing (SEDAR 2009), while in the US Atlantic it is both overfished and undergoing overfishing (SEDAR 2010). If it is true that indirect interactions effectively drive all trophic cascades as Schmitz et al. (2004) have argued, and that failures in fisheries management ultimately stem from an under-appreciation of the complexity and magnitude of these interactions (Travis et al. 2014), then fisheries-induced declines in the Red Grouper population will translate into fewer indirect interactions with largely unknown consequences.

Human-altered ecosystems may be more easily invaded by exotic species, and the invasion literature suggests that exotic species can more easily colonize altered habitats

115 compared to intact habitats (Sax and Brown 2000; Mack et al. 2000). Albins and Hixon’s (2013)

description of a “worst-case scenario” for lionfish in the western Atlantic highlights the need for intact predator communities necessary to ameliorate the effects of the lionfish invasion. The

present study provides the first experimental evidence of this effect and begins to shed light on

the mechanisms by which native predators may lessen the negative effects of this voracious

exotic invader.

116 Table 4.1. Physical characteristics and treatment assignments for 18 solution holes located in Florida Bay that were used to test Red Grouper effects in 2012 and 2013. Treatment codes for 2012: “RG-” Red Grouper removed; “RG+” Red Grouper present. Treatment codes for 2013: “NP” no predators present; “LO” lionfish only, Red Grouper removed; “RG” Red Grouper only, no lionfish; “L+RG” both a lionfish and a Red Grouper were present. Holes # 2, 2A, and 3 were discovered during diver surveys in 2013, so no data is available for 2012. For all solution holes, the area and maximum depth values are shown as measured in 2013.

Treatment Group Hole # Area (m2) Max depth (cm) 2012 2013 1 -- NP 2.18 33 2 n.d. LO 1.69 77 2A n.d. NP 2.94 40 3 n.d. RG 4.06 46 4 RG- -- 6.99 52 5 RG+ LO 6.31 46 6 RG+ L+RG 6.08 50 7 -- NP 5.45 26 8 -- RG 2.52 47 9 RG- L+RG 6.32 44 11 -- L+RG 3.69 35 12 -- LO 4.73 49 13 RG+ -- 5.86 48 17 RG+ NP 5.01 55 18 RG- RG 3.98 39 23 RG- RG 2.72 59 24 RG+ L+RG 5.01 46 25 RG- LO 4.47 33

117 150 RG+ 125 RG-

100

75

50 Recruit abundance 25

0 0 1 2 3 4 Week

Figure 4.1. Abundance of juvenile coral reef fish recruits over time at solution holes in Florida Bay with (closed circles; N = 5) and without (open circles; N = 5) Red Grouper. Values presented are means ± SE for each group. Data was collected during diver surveys of solution hole fish communities during the summer of 2012.

150 Lionfish + Red Grouper No predator 125 Lionfish only Red Grouper only A 100

75

50 AB

Recruit Abundance Recruit B 25 B 0 0 1 2 3 4 5 6 Week

Figure 4.2. Abundance of juvenile coral reef fish recruits over time at solution holes in Florida Bay from four experimental predator treatments in 2013. Values presented are means ± SE for each group; the sample size for each group was 4. Letters at the far right indicate results of pairwise comparisons performed on recruit abundance among treatments at the final census (matching letters indicate a p-value > 0.05 based on simultaneous tests performed on the best fit linear mixed-effects model).

118 Table 4.2. Estimated means of recruit abundance (± standard errors [SE]) and p-values resulting from pairwise t-tests on each of the six a priori contrasts based on the four predator treatment for each of the five community response variables at the end of the 6-week Red Grouper and lionfish experiment conducted in Florida Bay in 2013. The sample size for each predator treatment was 4. P-values less than 0.10 are indicated in bold; p-values less than 0.05 are indicated in italics.

Response Treatment Estimate ± SE Contrast p-value Abundance NP v. L 0.037

No Predator 43.8 ± 17.2 NP v. RG 0.050 Lionfish 8.25 ± 8.25 NP v. L + RG 0.413 Red Grouper 111 ± 35.4 L v. RG 0.023 Lionfish + Red Grouper 50.8 ± 15.2 L v. L + RG 0.027 RG v. L + RG 0.059 Species Richness NP v. L 0.010 (H0) No Predator 4.50 ± 0.866 NP v. RG 0.148 Lionfish 1.00 ± 1.00 NP v. L + RG 0.296 Red Grouper 3.25 ± 0.629 L v. RG 0.014 Lionfish + Red Grouper 5.75 ± 0.629 L v. L + RG < 0.001 RG v. L + RG 0.031 Shannon Diversity NP v. L 0.059 (H ) 1 No Predator 2.78 ± 0.168 NP v. RG 0.171 Lionfish 0.864 ± 0.864 NP v. L + RG 0.448 Red Grouper 2.30 ± 0.41 L v. RG 0.104 Lionfish + Red Grouper 3.24 ± 0.708 L v. L + RG 0.033 RG v. L + RG 0.208 Simpson Diversity NP v. L 0.077 (H2) No Predator 2.25 ± 0.0823 NP v. RG 0.722 Lionfish 0.776 ± 0.776 NP v. L + RG 0.470 Red Grouper 2.09 ± 0.409 L v. RG 0.097 Lionfish + Red Grouper 2.55 ± 0.347 L v. L + RG 0.053 RG v. L + RG 0.427 Evenness (H /H ) 2 1 NP v. L 0.080 No Predator 0.816 ± 0.032 NP v. RG 0.083 Lionfish 0.224 ± 0.224 NP v. L + RG 0.663 Red Grouper 0.900 ± 0.0214 L v. RG 0.058 Lionfish + Red Grouper 0.795 ± 0.032 L v. L + RG 0.086 RG v. L + RG 0.042

119 Table 4.3. Model selection results of linear mixed model analysis of the Red Grouper exclusion experiment conducted in Florida Bay in 2012. Likelihood Ratio Test (LRT) results are shown for the test of including a treatment by time interaction term; if the interaction did not significantly improve the model based on the LRT, then treatment and week were tested independently by dropping the term and testing that model against one with both factors. Test results of including a variance by treatment and autocorrelation structure in models were determined based on AIC scores as adding these structures caused models to be non-nested. Significant effects, shown in bold, are those whose inclusion improved the model fit; the final optimal model included all factors and structures in bold.

Response LRT Results Variance Autocorrelation AIC Variable Variable L-ratio df p-value Structure Structure

equal none 462.3 treat*week 15.6 1 < 0.001 treatment none 464.1 N treatment ------equal AR(1) 461.7 week ------treatment AR(1) 470.1 equal none 152.4 treat*week 9.64 1 0.002 treatment none 151.3 H treatment ------0 equal AR(1) 154.5 week ------treatment AR(1) 153.3

equal none 103.9 treat*week 2.14 1 0.143 treatment none 104.3 H treatment 1.95 1 0.162 1 equal AR(1) 105.2 week 4.47 1 0.035 treatment AR(1) 106.1

equal none 77.0 treat*week 2.14 1 0.143 treatment none 77.4 H treatment 0.466 1 0.495 2 equal AR(1) 78.2 week < 0.001 1 0.986 treatment AR(1) 78.8

equal none -106.1 treat*week 5.04 1 0.025 treatment none -104.1 Evenness treatment ------equal AR(1) -104.8 week ------treatment AR(1) -102.9

120 Table 4.4. Model selection results of linear mixed model analysis of the Red Grouper and lionfish experiment conducted in Florida Bay in 2013. Likelihood Ratio Test (LRT) results are shown for the test of including a treatment by time interaction term; if the interaction did not significantly improve the model based on the LRT, then treatment and week were tested independently by dropping the term and testing that model against one with both factors. Test results of including a variance by treatment and autocorrelation structure in models were determined based on AIC scores as adding these structures caused models to be non-nested. Significant effects, shown in bold, are those whose inclusion improved the model fit; the final optimal model included all factors and structures in bold.

Response LRT Results Variance Autocorrelation AIC Variable Variable L-ratio df p-value Structure Structure

equal none 1083 treat*week 31.0 18 0.029 treatment none 994.2 N treatment ------equal AR(1) 1066 week ------treatment AR(1) 952.4 equal none 414.8 treat*week 22.8 18 0.200 treatment none 418.4 H treatment 17.8 3 < 0.001 0 equal AR(1) 402.3 week 25.6 6 < 0.001 treatment AR(1) 403.8

equal none 338.7 treat*week 16.8 18 0.53 treatment none 342.5 H treatment 15.7 3 0.001 1 equal AR(1) 340.5 week 21.6 6 0.001 treatment AR(1) 343.6

equal none 317.1 treat*week 18.2 18 0.44 treatment none 320.0 H treatment 16.1 3 0.001 2 equal AR(1) 318.7 week 16.1 6 0.013 treatment AR(1) 322.0

equal none 104.4 treat*week 29.5 18 0.04 treatment none 92.1 Evenness treatment ------equal AR(1) 105.4 week ------treatment AR(1) 93.1

121 A. B. 3.0 6

2.0 4

2 1.0 Species Richness Shannon Diversity

p = 0.011 p = 0.160 0 0.0 RED NONE RED NONE

1.0 C. D.

2.0

0.9

1.0 0.8 Hill's Evenness Simpson's Diversity Simpson's

p = 0.373 p = 0.019 0.0 0.7 RED NONE RED NONE

Figure. 4.3. Species richness (A), diversity (Shannon [B] and Simpson [C]), and Hill’s evenness (D) after four weeks with (closed circles; N = 5) or without (open circles; N = 5) Red Grouper in 2012 solution hole recruitment experiment. P-values shown are results of t-tests performed on week-4 data.

122

A. C B. A 6 4.0 B,C A 3.0 A,B 4 A

2.0 B D 2 Species Richness Shannon Diversity 1.0

0 0.0 RED LION NONE BOTH RED LION NONE BOTH

C. D. 3.0 A 1.0 A A C C A 0.8 2.0 B 0.6 B 0.4 1.0 Hill's Evenness Simpson's Diversity Simpson's 0.2

0.0 0.0 RED LION NONE BOTH RED LION NONE BOTH

Figure. 4.4. Species richness (A.), Shannon (B.) and Simpson’s diversity (C.), and Hill’s evenness (D.) of juvenile reef fish recruits after six weeks with Red Grouper alone (closed circles; N = 4), Lionfish alone (triangles; N = 4), neither Red Grouper or Lionfish (open circles; N = 4), or both Red Grouper and Lionfish (diamonds; N = 4). Data was collected in 2013 from solution holes in Florida Bay. Letters indicate significant differences between groups after six weeks based on pairwise t-tests performed on the best-fit linear mixed model.

123 Table 4.5. Relative change in mean recruit abundance by species for the 2012 Red Grouper exclusion experiment. The abundance of each species found during the final (week #4) diver surveys is shown as the Control value. The mean effect of Red Grouper is shown for each species and for the total abundance of juvenile reef fish recruits.

Control Red Grouper Effect Species Common name NRG- (NRG+ -NRG-) Haemulon plumierii White Grunt 23.6 60.4 Haemulon spp. UID grunt 0.0 13.2 Haemulon flavolineatum French Grunt 11.4 0.8 Pareques acuminatus Highhat 0.2 0.6 Acanthurus coeruleus Blue Tang 0.0 0.4 Anisotremus virginicus Porkfish 0.0 0.2 Lutjanus synagris Lane Snapper 0.0 0.2 Ocyurus chrysurus Yellowtail Snapper 0.0 0.2 Scarus iseri Striped Parrotfish 0.0 0.2 Thalassoma bifasciatum Bluehead Wrasse 0.0 0.2 Pomacanthus arcuatus Gray Angelfish 0.8 -0.4 Total 36.0 +76.0 % change -- +211%

124

Figure 4.5. NMDS ordination of juvenile reef fish communities associated with experimental solution holes with (N = 5; closed circles), and without (N = 5; open circles) Red Grouper after 4-weeks of Red Grouper presence (“Control (RG+)”) or when Red Grouper were experimentally removed from solution holes (“Removal (RG-)”). Ellipses represent the standard deviation of all points for each group.

125

Table 4.6. Relative change in mean recruit abundance by species for the 2013 lionfish experiment. The abundance of each species found during the final (week #6) diver surveys for the No Predator treatment is shown as the Control value. The mean effect of Red Grouper, Lionfish, both predators together, and the BMII effect are shown for each species and for the total abundance of juvenile reef fish recruits.

Control Red Grouper Effect Lionfish effect Combined effect BMII effect Species Common name NNo Predator (NRG - NNP) (NLO – NNP) (NBoth – NNP) (NBoth – NLO) Haemulon plumierii White Grunt 17.3 47.8 -13.5 8.5 22.0 Haemulon flavolineatum French Grunt 11.3 14.5 -9.0 2.75 11.8 Haemulon spp. UID grunt 11 7.75 -9.5 -4.5 5 Anisotremus virginicus Porkfish 0.25 0.25 -0.25 0.5 0.75 Pomacanthus arcuatus Gray Angelfish 0.0 0.0 0.0 0.75 0.75 Holacanthus ciliaris Queen Angelfish 0.25 -0.25 -0.25 0.0 0.25 Haemulon parra Sailors Choice 1.0 -0.5 -1.0 0.0 1.0 Lutjanus synagris Lane Snapper 0.5 -0.5 -0.5 -0.5 0.0 Pareques acuminatus Highhat 2 -1.25 -1.25 -1.0 0.25

Total 43.5 +67.8 -35.3 +6.5 +41.8 % change -- +156% -81.1% +14.9% +96.1%

126

Figure 4.6. NMDS ordination of juvenile reef fish communities (abundance by species) associated with experimental solution holes at the end of the 6-week experiment conducted in Florida Bay during June – July, 2013. Ellipses show the standard deviation of all points for each predator treatment group: both Lionfish and Red Grouper (closed circles; N = 4); neither lionfish nor Red Grouper (open circles; N = 4); Lionfish Only (triangles; N = 4); and Red Grouper Only (diamonds; N = 4). There is no ellipse for the Lionfish Only predator treatment as 3 of the 4 juvenile fish communities were non-existent (n = 0) at the end of the experiment; these sites are represented by the single point located at (-1.33, 0.0217).

127 8.0

6.0

4.0

Effect size Effect 2.0

0.0

-2.0 CELion TIIRG,2012 TIIRG,2013 BMII

Effect type

Figure 4.7. Bootstrapped mean effect sizes with 95% confidence intervals for the consumptive effect of a single lionfish (CELion), the total indirect interactions of a single Red Grouper (TIIRG) in 2012 and 2013, and the estimated behaviorally-mediated indirect interaction (BMII) between Red Grouper and lionfish in terms of the abundance of solution-hole associated juvenile reef fish recruits.

128 CHAPTER FIVE

CONCLUSIONS

Engineer species have complex ecological effects through the combination of direct and indirect interactions they have with species that colonize engineered habitats (Jones et al. 2010).

My results suggest that for Red Grouper, a habitat engineer, the sum of these effects are mostly positive for the faunal communities that inhabit Florida Bay solution holes, and that their effects manifest through both direct and indirect pathways.

At the community-level, Red Grouper had an overall positive effect on both the abundance and diversity of solution-hole associated faunal communities. Red Grouper made holes deeper holes over time, and deeper holes were associated with more abundant and diverse faunal communities, suggesting that Red Grouper the positive effects on communities are at least partially mediated through changes in habitat. A short term experiment indicated that communities at solution holes with Red Grouper were more abundant and diverse compared to communities at solution holes where Red Grouper were excluded. However, the distribution of interaction strengths between Red Grouper and community members indicated that most species have no direct interaction, while just a few have strong, positive direct interactions with Red

Grouper. Together the results described in Chapter Two suggest that habitat-mediated indirect effects of Red Grouper may be more important for regulating the diversity of communities, while the indirect effects mediated through other species regulate the abundance of communities.

One of the species that had a strong positive interaction with Red Grouper, and also one that was responsible for much of the difference in diversity between solution holes with and without Red Grouper, was the Caribbean Spiny Lobster. Overall, lobsters were more abundant in solution holes when Red Grouper were also present, despite the fact that Red Grouper are lobster

129 predators. However this interaction changed through lobster ontogeny: juvenile lobsters do worse

with Red Grouper while sub-adult and adult lobsters do better with Red Grouper. The results of

measuring the survival of lobsters across size-classes suggested that this pattern was driven by

predation: small lobsters had a relatively high predation risk compared to larger lobsters. The

tethering experiment results also suggested that lobsters enter a size refuge where predation risk

declines rapidly with increasing size. The casita experiment results suggested that juvenile and

sub-adult lobsters do not appear to avoid Red Grouper occupied dens, as there was no evidence of avoidance of the casitas with a caged Red Grouper. Together these results suggest that

juvenile and smaller sub-adult lobsters below the size refuge, have a predator-prey interaction

with Red Grouper, while large sub-adult and adult lobsters in the size refuge have an indirect

habitat-mediated interaction with Red Grouper.

The varying interaction across lobster ontogeny may have interesting population-level

effects for lobsters: if the positive indirect effect of Red Grouper on adult lobsters is larger than

the negative direct effect on juvenile lobsters, then overall lobsters would benefit from greater

Red Grouper abundance. More Red Grouper would create more lobster habitat which may

decrease intra-species competition, and increase lobster growth rates which would decrease the

time until lobsters enter a size refuge. For Red Grouper, this could be an example of an indirect

mutualism: by excavating solution holes Red Grouper increase the amount of habitat for adult

lobsters, which, being gregarious, may attract smaller lobsters to solution holes where they are

consumed by Red Grouper. This is a potential example of a positive indirect feedback of

engineering as predicted by Jones et al. (2010). Further studies of this interaction, including

better estimates of predation risk face by lobsters and consumption rates by Red Grouper are

required to confirm this mutualism. Population modeling efforts could reveal interesting

130 predator-prey driven dynamics of lobsters and Red Grouper in Florida Bay. These studies would

be beneficial given that both of these species support economically significant fisheries in

Florida and given recent calls for fisheries managers to do better at including indirect interactions in management (Travis et al. 2014).

Juvenile coral reef fishes were the other group of species to have strong positive effects with Red Grouper. Juvenile reef fishes, primarily post-settlement recruits, were more abundant at

solution holes with Red Grouper. I hypothesized that this effect was primarily driven by changes

in behavior among solution-hole associated piscivores because Red Grouper are relatively large

and exhibit territorial displays towards both conspecifics and occasionally other solution-hole

associated fishes. These behaviors may disrupt piscivory enough to result in higher survival of

recruits with Red Grouper. Invasive lionfish were used as a proxy for native piscivores and the

predicted effects were confirmed: recruits were most abundant with Red Grouper alone, least

abundant with lionfish alone, and at intermediate abundance when lionfish were in solution holes

with Red Grouper. The experimental results suggest that Red Grouper have a behaviorally-

mediated indirect interaction with recruits via piscivores. These results also confirmed that

lionfish have strong negative effects on the abundance and diversity of native reef fishes in

Florida Bay, just as they have elsewhere in the invaded range. Red Grouper appeared to

ameliorate some of these negative effects, supporting the predictions of others

An untested assumption that I made for the experiments described in Chapter Four was

that recruitment of reef fishes was equivalent among holes with and without Red Grouper. Some

recent research suggests that chemical cues may have strong effects on reef fish settlement

(Dixson et al. 2014), and Red Grouper excavation behaviors may provide a source for attractant

cues. Testing this assumption would be relatively easy and would contribute to the understanding

131 of the mechanisms that drive indirect interactions between Red Grouper and reef fishes. Further studies into the specific behavioral interactions that solution-hole associated piscivores have with

Red Grouper would also be beneficial to better understand the mechanisms driving these effects.

These results are especially significant in light of the rapid expansion of lionfish throughout the

Caribbean and western Atlantic. Red Grouper reduced the negative effects of the lionfish when they were present together suggesting that native predators that compete for space or that represent a predation threat for lionfish can ameliorate the effects of the invasion. Even without a complete description of the mechanisms, it is clear that Red Grouper regulate the abundance of reef fishes in Florida Bay, with or without lionfish.

All together the results of my dissertation suggest that Red Grouper have important effects on the ecology of Florida Bay. Through their excavating activities, territorial behaviors, and role as predators, Red Grouper increase the amount of habitat used by other species, and have direct interactions that modify survival and demographic rates of others. Most importantly,

Red Grouper are a perfect example of why the full complement of species interactions must be taken into account when managing fisheries. Solution holes are relatively rare in Florida Bay, and the effects of Red Grouper on associated communities were localized to the immediate vicinity around the holes. However, Red Grouper affected species colonized solution holes temporarily as they moved throughout the bay or onto adjacent reef habitats. If these effects translate to the population-level, one may have a strong argument that Red Grouper do not just dig hole but in fact engineer ecosystems.

132 APPENDIX A

EXTRA FIGURES AND TABLES

Table A.1. Fish species observed at solution holes used for observational (OBS) and experimental (EXP) components of the study conducted in Florida Bay from 2010 – 2013. Functional group classifications were based on reported diet information for the Feeding group, and the location of individuals in relation to solution holes as observed during diver surveys for the Habitat group. BA = herbivores (consume primarily benthic algae); CL = cleaners (consume ectoparasites); INV = invertevores; PL = plantktivores; PV = piscivores; ZB = benthivores (consume primarily small benthic invertebrates); DEM = demersal fishes primarily observed inside solution holes; MILL = milling behavior, where the fish was found in the water-column above and around solution holes; TRANS = transient fishes observed visiting solution holes.

Functional Group Study component Family Species Common name Feeding Habitat OBS EXP

Acanthuridae Acanthurus bahianus Ocean Surgeonfish BA MILL X Acanthurus chirurgus Doctorfish BA MILL X X Acanthurus coeruleus Blue Tang BA MILL X X

Apogonidae Apogon binotatus Barred Cardinalfish PL DEM X X Apogon maculatus Flamefish ZB DEM X X Apogon pseudomaculatus Twospot Cardinalfish PL DEM X

Balistidae Balistes capriscus Gray Triggerfish ZB MILL X

Batrachoididae Opsanus beta Gulf Toadfish PV DEM X

Callionymidae Diplogrammus pauciradiatus Spotted Dragonet ZB MILL X

Carangidae Caranx ruber Bar Jack PV TRANS X

Chaetodontidae Chaetodon ocellatus Spotfin Butterflyfish ZB MILL X X Chaetodon sedentarius Reef Butterflyfish ZB MILL X

Gerreidae Eucinostomus melanopterus Flagfin Mojarra ZB TRANS X X

133 Table A.1 – continued

Functional Group Study component Family Species Common name Feeding Habitat OBS EXP

Gobiidae Bathygobius soporator Frillfin Goby ZB DEM Coryphopterus glaucofraenum Bridled Goby BA DEM X X Elacatinus oceanops Neon Goby CL DEM X X Gobisoma macrodon Tiger Goby ZB DEM X

Haemulidae Anisotremus virginicus Porkfish CL/ZB MILL X X Haemulon chrysargyreum Smallmouth Grunt ZB MILL X Haemulon flavolineatum French Grunt ZB MILL X X Haemulon melanurum Cottonwick ZB MILL X Haemulon parra Sailors Choice ZB MILL X Haemulon plumierii White Grunt ZB MILL X X Haemulon sciurus Bluestriped Grunt ZB MILL X X

Labridae Halichoeres bivittatus Slippery Dick ZB TRANS X Halichoeres radiates Puddingwife ZB TRANS X Lachnolaimus maximus Hogfish ZB TRANS X X Thalassoma bifasciatum Bluehead Wrasse CL/PL TRANS X X

Lutjanidae Lutjanus griseus Gray Snapper ZB TRANS X X Lutjanus synagris Lane Snapper ZB TRANS X X Ocyurus chrysurus Yellowtail Snapper PV TRANS X X

Mullidae Pseudupeneus maculatus Spotted Goatfish ZB TRANS X X

Muraenidae Gymnothorax funebris Green Moray INV/PV DEM X Gymnothorax moringa Spotted Moray INV/PV DEM X Gymnothorax vicinus Purplemouth Moray INV/PV DEM X X

Pomacanthidae Holacanthus bermudensis Blue Angelfish ZB MILL X X Holacanthus ciliaris Queen Angelfish ZB MILL X X Pomacanthus arcuatus Gray Angelfish ZB MILL X X

Pomacanthus paru French Angelfish CL/ZB MILL X X

Pomacentridae Abudefdef saxatilis Sergeant Major ZB MILL X Stegastes variabilis Cocoa Damselfish BA MILL X

134 Table A.1 – continued

Functional Group Study component Family Species Common name Feeding Habitat OBS EXP

Pomacentridae Abudefdef saxatilis Sergeant Major ZB MILL X Stegastes variabilis Cocoa Damselfish BA MILL X

Rhincodontidae Ginglymostoma cirratum Nurse Shark INV DEM X X

Scaridae Scarus coeruleus Blue Parrotfish BA TRANS X X Scarus iseri Striped Parrotfish BA TRANS X X Sparisoma aurofrenatum Redband Parrotfish BA TRANS X Sparisoma viride Stoplight Parrotfish BA TRANS X

Sciaenidae Equetus lanceolatus Jacknife Fish ZB DEM X Pareques acuminatus Highhat ZB DEM X X

Scorpaenidae Pterois miles/volitans Lionfish PV DEM X X Scorpaena plumieri Spotted Scorpionfish PV DEM X X

Serranidae Diplectrum formosum Sand Perch ZB TRANS X X Epinephelus itajara Goliath Grouper INV DEM X Epinephelus morio Red Grouper INV DEM X X Hypoplectrus puella Barred Hamlet ZB MILL X X Mycteroperca bonaci Black Grouper PV MILL X X

Sparidae Calamus calamus Saucereye Porgy ZB TRANS X X Lagodon rhomboides Pinfish ZB TRANS X

Tetraodontidae Sphoeroides spengleri Bandtail Puffer ZB TRANS X

135 Table A.2. Motile macroinvertebrate species observed at solution holes used for observational (OBS) and experimental (EXP) components of the study conducted in Florida Bay from 2010 – 2013. Functional group classifications were based on reports of species consumed by Red Grouper (RG Diet) or as consuming ectoparasites (CL) from the literature.

Functional Group Study Component Phylum Class Order Species Common name RG Diet CL OBS EXP

Arthropoda Decapoda Hippolytidae Lysmata spp. Peppermint Shrimp N Y X X Thor amboinensis Squat Anemone Shrimp N N X

Majidae Mithrax forceps Red-ridged Clinging Crab Y N X Mithrax sculptus Green Clinging Crab Y N X Mithrax spinosissimus Channel Clinging Crab Y N X X Stenorhynchus seticornis Yellowline Arrow Crab N N X X

Palaemonidae Ancylomenes pedersoni Pederson’s Cleaner Shrimp N Y X X Periclimines yucatanicus Spotted Cleaner Shrimp N Y X X

Palinuridae Panulirus argus Caribbean Spiny Lobster Y N X X

Portunidae Portunus spinimanus Blotched Swimming Crab Y N X

Scyllaridae Scyllarides aequinoctialis Spanish Lobster Y N X Scyllarides nodifer Ridged Slipper Lobster Y N X

Stenopodidae Stenopus hispidus Banded Coral Shrimp N N X

Xanthidae Menippe mercenaria Florida Stone Crab N N X X

Echinodermata Echinoidea Clypeasteroida Clypeaster roasceus Inflated Sea Biscuit N N X X

Mollusca Cephalopoda Octopoda Octopus briareus Caribbean Reef Octopus Y N X

Gastropoda Cypraea cervus Atlantic Deer Cowrie N N X

136 25 P/A R² = 0.9049

W.C. 20

15 R² = 0.5777

SIMPER rank SIMPER 10

5

0 0 5 10 15 20 25 Rank abudnance

Figure A.1. Plot of SIMPER analysis results of whole community data (closed circles, “W.C.”) and presence/absence data (open circles, “P/A”) against rank abundance from diver observations.

Table A.3. Results of PERMANOVA analysis of species composition data for observational data of faunal communities at solution holes; “treat” refers to the Red Grouper occupancy effect. Bold p-values indicate significant effects.

PERMANOVA df SS MS Pseudo-F R2 P(perm)

treat 1 3.498 3.498 13.555 0.109 0.001

site 2 2.865 1.432 5.550 0.0892 0.001

year 1 1.487 1.487 5.764 0.0462 0.001

Residuals 94 24.258 0.258 0.7556

Total 98 32.108 1.0000

137

Figure A.2. Undergraduate intern measuring lobsters in situ at a solution hole in Florida Bay, June 2011.

138

Figure A.3. Images of casitas built to test lobster avoidance of Red Grouper. Clockwise from top left: casita frame during construction in 2012; detail of casita frame during construction showing placement of rebar supports; casita with cage and Red Grouper; casita in 2013 after one year on the bottom.

139

Figure A.4. Evidence of triggerfish predation events on tethered lobsters in 2013.

140 APPENDIX B

IACUC APPROVAL LETTER

ANIMAL CARE AND USE COMMITTEE [ACUC] 101 BIOMEDICAL RESEARCH FACILITY TALLAHASSEE, FL 32306-4341 TELEPHONE: 644-4262 FAX: 644-5570 MAIL CODE: 4341

November 4, 2015

The Graduate School Florida State University

To Whom It May Concern:

Concerning the thesis/dissertation submitted to the Graduate School by:

Graduate Student: Robert D. Ellis Thesis/Dissertation Ecological effects of Red Grouper (Epinephelus morio) in Title: Florida Bay Department: BIOLOGICAL SCIENCE

Major Professor: Felicia Coleman

141

The above named graduate student has provided assurance to the FSU Animal Care and Use Committee that all animal procedures utilized in the work resulting in this thesis/dissertation are described in FSU ACUC Protocol(s):

Protocol Title Date ACUC Number Approval Habitat and demographics of fishes of the Gulf of Mexico March 18, 2011 1106 and the South Atlantic Bight. Habitat and demographics of fishes of the Gulf of Mexico April 16, 2014 1411 and the South Atlantic Bight.

The Animal Care and Use Committee has confirmed that this student was included as a project member during the period covering their thesis/dissertation work. This institution has an Animal Welfare Assurance on file with the Office for Laboratory Animal Welfare. The Assurance Number is A3854- 01.

Sincerely, Kathleen Harper ACUC Attending Veterinarian FSU Animal Care and Use Committee

KMH/kjj cc: Robert D. Ellis Dr. Felicia Coleman Dr. Christopher C. Koenig

142 REFERENCES

Abrams, P.A., B.A. Menge, G.G. Mittelbach, D.A. Spiller, and P. Yodzis. 1996. The role of indirect effects in food webs. In Food Webs. Polis, G.A. and K. Winemiller (eds.). Springer USA. 472 pp.

Abrams, P.A. 1995. Implications of dynamically variable traits for identifying, classifying, and measuring direct and indirect effects in ecological communities. The American Naturalist. 146: 112-134.

Albins, M.A. 2013. Effects of invasive Pacific red lionfish Pterois volitans versus a native predator on Bahamian coral-reef fish communities. Biological Invasions. 15: 29-43.

Albins, M.A. 2015. Invasive Pacific lionfish Pterois volitans reduce abundance and species richness of native Bahamian coral-reef fishes. Marine Ecology Progress Series. 522: 231- 243. Albins, M.A. and M.A. Hixon. 2008. Invasive Indo-Pacific lionfish Pterois volitans reduce recruitment of Atlantic coral-reef fishes. Marine Ecology Progress Series. 367: 233-238.

Albins, M.A. and M.A. Hixon. 2013. Worst case scenario: potential long-term effects of invasive predatory lionfish (Pterois volitans) on Atlantic and Caribbean coral-reef communities. Environmental Biology of Fishes. 96: 1151-1157.

Almany, G.R. and M.S. Webster. 2006. The predation gauntlet: early post-settlement mortality in reef fishes. Coral Reefs. 25: 19-22.

Anderson, M.J. 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecology. 26: 32-46.

Arce, A.M., W. Aguilar-Davila, E. Sosa-Cordero, and J.F. Caddy. 1997. Artificial shelters (casitas) as habitats for juvenile spiny lobsters Panulirus argus in the Mexican Caribbean. Marine Ecology Progress Series. 158: 217-224.

Aronson, R.B. and K.L. Heck Jr. 1995. Tethering experiments and hypothesis testing in ecology. Marine Ecology Progress Series. 121: 307-309.

Ault, J.S., S.G. Smith, J.A. Bohnsack, J. Luo, D.E. Harper, and D.B. McClellan. 2006. Building sustainable fisheries in Florida’s coral reef ecosystem: positive signs in the Dry Tortugas. Bulletin of Marine Science. 78: 633-654.

Ault, J.S., S.G. Smith, J.A. Bohnsack, J. Luo, N. Zurcher, D.B. McClellan, T.A. Ziegler, D.E. Hallac, M. Patterson, M.W. Feeley, B.I. Ruttenberg, J. Hunt, D. Kimball, and B. Causey. 2013. Assessing coral reef fish population and community changes in response to marine reserves in the Dry Tortugas, Florida USA. Fisheries Research. 144: 28-37.

143 Bangert, R.K. and C.N. Slobodchikoff. 2002. The Gunnison’s prairie dog structures a high desert grassland landscape as a keystone engineer. Journal of Arid Environments. 46: 357-369.

Bates, D. M. Maechler, B. Bolker, and S. Walker. 2015. Lme4: Linear mixed-effects models using Eigen and S4. R package version 1.1-8.

Behringer, D.C. and M.J. Butler IV. 2010. Disease avoidance influences shelter use and predation in Caribbean spiny lobster. Behavioral Ecology and Sociobiology. 64: 747-755.

Beschta, R.L. and W.J. Ripple. 2009. Large predators and trophic cascades in terrestrial ecosystems of the western United States. Biological Conservations. 142: 2401-2414.

Berger, D.K. and M.J. Butler IV. 2001. Octopuses influence den selection by juvenile Caribbean spiny lobster. Marine and Freshwater Research. 52: 1049-1053.

Berlow, E.L., S.A. Navarrete, C.J. Briggs, M.E. Power, and B.A. Menge. 1999. Quantifying variation in the strengths of species interactions. Ecology. 80: 2206-2224.

Bertelsen, R.D. and C. Cox. 2001. Sanctuary roles in population and reproductive dynamics of Caribbean spiny lobster, in “Spatial Processes and Management of Marine Populations” (Kruse, G.H, N. Bez, A. Booth, M.W. Dorn, S. Hills, R.N. Lipcius, D. Pelletier, C. Roy, S.J. Smith, and D. Witherells, Eds.). pp 591-605. University of Alaska Sea Grant.

Beukers, J.S. and G.P. Jones. 1997. Habitat complexity modifies the impact of piscivores on a coral reef fish population. Oceologia. 114: 50-59.

Biggs, C.R. and J.D. Olden. Multi-scale habitat occupancy of invasive lionfish (Pterois volitans) in coral reef environments of Roatan, Honduras. Aquatic Invasions. 6: 347-353,

Boogert, N.J., D.M. Paterson, and K.L. Laland. 2006. The implications of niche construction and ecosystem engineering for conservation biology. Bioscience. 56: 570-578.

Bouwma, P.E. 2006. Aspects of antipredation in Panulirus argus and Panulirus guttatus: behavior, morphology, and ontogeny. PhD Dissertation, Florida State University, Tallahassee, Florida, United States. 114 pp.

Brashares, J.S., L.R. Prugh, C.J. Stoner, and C.W. Epps. 2010. Ecological and conservation implications of mesopredator release. In Trophic cascades: predators, prey, and the changing dynamics of nature. Terborgh, J., and J.A. Estes (eds.). Island Press, Washington, USA. pp. 221-240.

Breitburg, D.L., B.C. Crump, J.O. Dabiri, and C.L. Gallegos. 2010. Ecosystem engineers in the pelagic realm: alteration of habitat by species ranging from microbes to jellyfish. Integrative and Comparative Biology. 50: 188-200.

144 Brooks, J.L. and S.I. Dodson. 1965. Predation, body size, and composition of plankton. Science. 150: 28-35.

Brulé, T. and L.G. Rodriguez Canche, 1993 Food habits of juvenile red groupers, Epinephelus morio (Valenciennes, 1828), from Campeche Bank, Yucatan, Mexico. Bulletin of Marine Science. 52:772-779.

Bullock, L.H. and G.B. Smith. 1991. Seabasses (Pisces: Serranidae). Memoirs of the Hourglass Cruises, Volume VIII; Part II. Florida Marine Research Institute, Department of Natural Resources, St. Petersburg, Florida, USA. 243 pp.

Burkenbusch, K., A.A. Rowden, and T.E. Meyers. 2007. Interactions between seagrasses and burrowing ghost shrimps and their influence on infaunal assemblages. Journal of Experimental Marine Biology and Ecology. 341: 70-84.

Butler IV, M.J., J.H. Hunt, W.F. Herrnkind, M.J. Childress, R. Bertelsen, W. Sharp, T. Matthews, J.M. Field, H.G. Marshall. 1995. Cascading disturbances in Florida Bay, USA: cyanobacteria blooms, sponge mortality, and implications for juvenile spiny lobsters Panulirus argus. Marine Ecology Progress Series. 129: 119-125.

Butler IV, M.J., A.B. MacDiarmid, J.D. Booth. 1999. The cause and consequence of ontogenetic changes in social aggregation in New Zealand spiny lobsters. Marine Ecology Progress Series. 188: 179-191.

Butler IV, M.J. and J.A. Lear. 2009. Habitat-based intraguild predation by Caribbean reef octopus Octopus briareus on juvenile Caribbean spiny lobster Panulirus argus. Marine Ecology Progress Series. 386: 115-122.

Carr, M.H. and M.A. Hixon. 1995. Predation effects on early post-settlement survivorship of coral-reef fishes. Marine Ecology Progress Series. 124: 31-42.

Caselle, J.E. 1999. Early post-settlement mortality in a coral reef fish and its effect on local population size. Ecological Monographs. 69: 177-194.

Castilla, J.C., N.A. Lagos, and M. Cerda. 2004. Marine ecosystem engineering by the alien ascidian Pyura praeputialis on a mid-intertidal rocky shore. Marine Ecology Progress Series. 268: 119-130.

Caswell. 1978. Predator-mediated coexistence: a nonequilibrium model. American Naturalist. 112: 127-154.

Catano, C.P. and I.J. Stout. 2015. Functional relationships reveal keystone effects of the gopher tortoise on vertebrate diversity in a longleaf pine savannah. Biodiversity and Conservation. 24: 1957-1974.

145 Ceballos, G., J. Pacheco, and R. List. 1999. Influence of prairie dogs (Cynomys ludovicianus) on habitat heterogeneity and mammalian diversity in Mexico. Journal of Arid Environments. 41: 161-172.

Chang, E.S. 1995. Physiological and biochemical changes during the molt cycle in decapod crustaceans: an overview. Journal of Experimental Marine Biology and Ecology. 193: 1- 14.

Chao, A., C. Chiu, and L. Jost. 2014. Unifying species diversity, phylogenetic diversity, functional diversity, and related similarity and differentiation measures through Hill numbers. Annual Review of Ecology, Evolution, and Systematics. 45: 297-324.

Charmantier, G., M. Charmantier-Daures, and D.E. Aiken. 1991. Metamorphosis in the lobster Homarus (Decapoda): a review. Journal of Crustacean Biology. 11: iii-495.

Childress, M. J. and W. F. Herrnkind. 1996. The ontogeny of social behavior among juvenile Caribbean spiny lobsters. Animal Behavior. 51: 675-687.

Childress, M. J. and W. F. Herrnkind. 2001. The guide effect influence on gregariousness of juvenile Caribbean spiny lobsters. Animal Behavior. 62: 465-472.

Cocheret de la Morinière, E., B.J.A. Pollux, I. Nagelkerken, and G. van der Velde. 2003. Diet shifts of Caribbean grunts (Haemulidae) and snappers (Lutjanidae) and the relation with nursery-to-coral reef migrations. Estuarine, Coastal and Shelf Science. 57: 1079-1089.

Coleman, F.C. and S.L. Williams. 2002. Overexploiting marine ecosystem engineers: potential consequences for biodiversity. Trends in Ecology and Evolution. 17: 40-44.

Coleman, F.C., C.C. Koenig, K.M. Scanlon, S. Heppell, S. Heppell, and M.W. Miller. 2010. Benthic habitat modification through excavation by red grouper, Epinephelus morio, in the northeastern Gulf of Mexico. Open Fish Science Journal. 3: 1-15.

Coleman, F.C., K.M. Scanlon, and C.C. Koenig. 2011. Groupers on the edge: Shelf edge spawning habitat in and around marine reserves in the northeastern Gulf of Mexico. The Professional Geographer. 63: 1-19.

Connell, J.H. 1975. Some mechanisms producing structure in natural communities: a model and evidence from field experiments, in “Ecology and Evolution of Communities” (Cody, M.L. and J.M. Diamond, Eds.). pp 460-490. Belknap Press, Cambridge.

Connell, J.H. 1978. Diversity in tropical rain forests and coral reefs. Science. 199: 1302-1310.

Côté, I.M., S.J. Green, J.A. Morris Jr., J.L. Akins, D. Steinke. 2013. Diet richness of invasive Indo-Pacific lionfish revealed by DNA barcoding. Marine Ecology Progress Series. 472: 249-256.

146 Côté, I.M., S.J. Green, and M.A. Hixon. 2013. Predatory fish invaders: insights from Indo- Pacific lionfish in the western Atlantic and Caribbean. Biological Conservation. 164: 50-. 61.

Cox, C. and J.H. Hunt. 2005. Change in size and abundance of Caribbean spiny lobsters Panulirus argus in a marine reserve in the Florida Keys National Marine Sanctuary, USA. Marine Ecology Progress Series. 204: 227-239.

Dayton, P.K., S.F. Thrush, M.T. Agardy, and R.J. Hoffman. 1995. Environmental effects of marine fishing. Aquatic Conservation: Marine and Freshwater Ecosystems. 5: 205-232. de Roos, A.M. and L. Persson. 2002. Size-dependent life-history traits promote catastrophic collapses of top predators. Proceedings of the National Academy of Sciences of the United States of America. 99: 12907-12912. de Roos, A.M., L. Persson, and E. McCauley. 2003. The influence of size-dependent life-history traits on the structure and dynamics of populations and communities. Ecology Letters. 6: 473-487.

Diemer, J.E. 1986. The ecology and management of the Gopher Tortoise of the southeastern United States. Herpetologica. 42: 125-133.

Dill, L.M., M.R. Heithaus, and C.J. Walters. 2003. Behaviorally mediated indirect interactions in marine communities and their conservation implications. Ecology. 84: 1151-1157.

Diller, J.L., T.K. Frazer, and C.A. Jacoby. 2014. Coping with the lionfish invasion: evidence that naïve, native predators can learn to help. Journal of Experimental Marine Biology and Ecology. 455: 45-49.

Dixson, D.L., D. Abrego, and M.E. Hay. 2014. Chemically mediated behavior of recruiting corals and fishes: A tipping point that may limit reef recovery. Science. 345: 892-897.

Doherty, P.J., D. M. Williams, and P.F. Sale. 1985. The adaptive significance of larval dispersal in coral reef fishes. Environmental Biology of Fishes. 12:81-90.

Doherty, P.J. 2002. Variable replenishment and the dynamics of reef fish populations, in “Coral Reef Fishes: Dynamics and Diversity in a Complex Ecosystem” (Sale, P.F., Ed.). pp 327- 357. Academic Press, San Diego.

Dulvy, N.K., R.P. Freckleton, and N.V.C. Polunin. 2004. Coral reef cascades and the indirect effects of predator removal by exploitation. Ecology Letters. 7: 410-416.

Eggleston, D.B., R.M. Lipcius, D.L. Miller, and L. Coba-Centina. 1990. Shelter scaling regulates survival of juvenile spiny lobster, Panulirus argus. Marine Ecology Progress Series. 62: 79-88.

147 Eggleston, D.B. and R.M. Lipcius. 1992. Shelter selection by spiny lobster under variable predation risk, social conditions, and shelter size. Ecology. 73: 992-1011.

Eggleston, D.B., R.N. Lipcius, and J.J. Grover. 1997. Predator and shelter-size effects on coral reef fish and spiny lobster prey. Marine Ecology Progress Series. 149: 43-59.

Elise, S., I. Urbina-Barreto, H. Boadas-Gil, M. Galindo-Vivas, and M. Kulbicki. 2014. No detectable effect of lionfish (Pterois volitans and P. miles) invasion on a healthy reef fish assemblage in Archipelago Los Roques National Park, Venezuela. Marine Biology. 162: 319-330.

Estes, J.A. and J.F. Palmisano. 1974. Sea otters: their role in structuring nearshore communities. Science. 185: 1058-1060.

Estes, J.A., N.S. Smith, and J.F. Palmisano. 1978. Sea otter predation and community organization in the Western Aleutian Islands, Alaska. Ecology. 59: 822-833.

Estes, J.A. and D.O. Duggins. 1995. Sea otters and kelp forests in Alaska: Generality and variation in a community ecological paradigm. Ecological Monographs. 65: 75-100.

Estes, J.A., M.T. Tinker, and J.L. Bodkin. 2010. Using ecological function to develop recovery criteria for depleted species: Sea otters and kelp forests in the Aleutian archipelago. Conservation Biology. 24: 852-860.

Fitzhugh, G.R., H.M. Lyon, W.T. Walling, C.F. Levins, and L.A. Lombardi-Carlson. 2006. An update of Gulf of Mexico red grouper reproductive parameters for SEDAR 12. National Marine Fisheries Service, Panama City Laboratory Contribution 06-14. 17 pp.

Fourqurean, J.W., M.J. Durako, M.O. Hall, L.N. Hefty. 2002. Seagrass distribution in South Florida: a multi-agency coordinated monitoring program. In: Porter, J.W. and K.G. Porter (eds) The Everglades, Florida Bay, and coral reefs of the Florida Keys: An ecosystem sourcebook, CRC Press, Boca Raton, FL. 1025 pp.

Gilpin, M.E. and J.M. Diamond. 1984. Are species co-occurences on islands non-random, and are null hypotheses useful in community ecology? In: Strong, D.R. et al. (eds) Ecological Communities, Princeton University Press, Princeton, New Jersey. 613 pp.

Grabowski, J.H. 2004. Habitat complexity disrupts predator-prey interactions but not the trophic cascade on oyster reefs. Ecology. 85: 995-1004.

Greenaway, P. 2003. Terrestrial adaptations in the Anomura (Crustacea: Decapoda), in “Biology of the Anomura” (Lemaitre, R., and C.C. Tudge, Eds.). Proceedings of a symposium at the Fifth International Crustacean Congress, Melbourne, Australia, 9–13 July 2001. Memoirs of Museum Victoria 60: 13–26.

148 Grabowski, J.H., A.R. Hughes, and D.L. Kimbro. 2008. Habitat complexity influences cascading effects of multiple predators. Ecology. 89: 3413-3422.

Green, S.J., J.L. Akins, A. Maljkovic, and I.M. Côté. 2013. Invasive lionfish drive Atlantic coral reef fish declines. PLoS ONE. 7: e32596.

Hedges, L.V. 1981. Distribution theory for Glass’s estimator of effect size and related estimators. Journal of Educational and Behavioral Statistics. 6: 107-128.

Heemstra, P.C. and J.E. Randall. 1993. Groupers of the world (family Serranidae, subfamily Epinephelinae). An annotated and illustrated catalogue of the grouper, rockcod, hind, coral grouper and lyretail species known to date. FAO Fisheries Synopsis. 125(16): 382 pp.

Heithaus, M.R. and L.M. Dill. 2002. Food availability and tiger shark predation risk influence bottlenose dolphin habitat use. Ecol. 83: 480-491.

Herrnkind, W.F. and M.J. Butler IV. 1986. Factors regulating postlarval settlement and juvenile microhabitat use by spiny lobsters Panulirus argus. Marine Ecology Progress Series. 34: 23-30.

Herrnkind, W.F., M.J. Butler IV, J.H. Hunt, and M. Childress. 1997. Role of physical refugia: implications from a mass sponge die-off in a lobster nursery in Florida. Marine and Freshwater Research. 48: 759-769.

Hill, M.O. 1973. Diversity and evenness: a unifying notation and its consequences. Ecology. 54: 427-432.

Hill, M.O. 1997. An evenness statistic based on the abundance-weighted variance of species proportions. Oikos. 79: 413-416.

Hixon, M.A. 1986. Fish predation and local prey diversity, in “Contemporary studies on fish feeding” (Simenstad, C.A. and G.M. Cailliet, Eds.). pp 235-257. Junk, Dordrecht, the Netherlands.

Hixon, M.A. 1991. Predation as a process structuring coral-reef fish communities. In The Ecology of Fishes on Coral Reefs, P.F. Sale (ed.). Academic Press; San Diego, California.

Hixon, M.A. and J.P. Beets. 1993. Predation, prey refuges, and the structure of coral-reef fish assemblages. Ecological Monographs. 63: 77-101.

Holt, R.D. 1987. Prey communities in patchy environments. Oikos. 50: 276-290.

Hovel, K.A. and M.S. Fonseca. 2005. Influence of seagrass landscape structure on the juvenile blue crab habitat-survival function. Marine Ecology Progress Series. 300: 179-191.

149 Hughes, A.R., K. Rooker, M. Murdock, and D.L. Kimbro. 2012. Predator cue and prey density interactively influence indirect effects on basal resources in intertidal oyster reefs. PLoS ONE. 7(9): e448d39.

Hunt, H.L. and R.E. Scheibling. 1997. Role of early post-settlement mortality in recruitment of benthic marine invertebrates. Marine Ecology Progress Series. 155: 269-301.

Jackson, D.R. and E.G. Milstrey. 1989. The fauna of gopher tortoise burrows. pp. 86-98 in “Gopher tortoise relocation symposium proceedings,” Diemer, J.E., D.R. Jackson, J.L. Landers, J.N. Layne, and D.A. Woods, eds. Florida Game and Fresh Water Fish Commission Nongame Wildlife Program Technical Report No. 5, Tallahassee, FL.

Jennings, S. and M.J. Kaiser. 1998. The effects of fishing on marine ecosystems. Advances in Marine Biology. 34: 201-352.

Johnson, D.W. 2007. Habitat complexity modifies post-settlement mortality and recruitment dynamics of a marine fish. Ecology. 88: 1716-1725.

Jones, C.G., J.H. Lawton, and M. Shachak. 1994. Organisms as ecosystem engineers. Oikos. 69: -373-386.

Jones, C.G., J.H. Lawton, and M. Shachak. 1997. Positive and negative effects of organisms as physical ecosystem engineers. Ecology. 78: 1946-1957.

Jones, C.G., J.L. Gutiérrez, J.E. Byers, J.A. Crooks, J.G. Lambrinos, and T.S. Talley. 2010. A framework for understanding physical ecosystem engineering by organisms. Oikos. 119: 1862-1869.

Jones, G.P., L. Santana, L.J. McCook, and M.I. McCormick. 2006. Resource use and impact of three herbivorous damselfishes on coral reef communities. Marine Ecology Progress Series. 328: 215–224.

Jud, Z.R. and C.A. Layman. 2012. Site fidelity and movement patterns of invasive lionfish, Pterois spp. in a Florida estuary. Journal of Experimental Marine Biology and Ecology. 414-415: 69-74.

Kaczor, S.A. and D.C. Hartnett. 1990. Gopher tortoise (Gopherus polyphemus) effects on soil and vegetation in a Florida sandhill community. American Midland Naturalist. 123:100– 111.

Kinlaw, A. and M. Grasmueck. 2011. Evidence for and geomorphologic consequences of a reptilian ecosystem engineer: the burrowing cascade initiated by the Gopher Tortoise. Geomorphology. 157-158: 108-121.

150 Lachaise A., A. Le Roux, M. Hubert, and R. Lafont. 1993. The molting gland of crustaceans: localization, activity, and endocrine control (a review). Journal of Crustacean Biology. 13:198–234.

Layman, C.A. and J.E. Allgeier. 2012. Characterizing trophic ecology of a generalist consumers: a case study of the invasive lionfish in The Bahamas. Marine Ecology Progress Series. 448: 131-141.

Lima, S.L. 1998. Nonlethal effects in the ecology of predator-prey interactions. Bioscience. 48: 25-34.

Lipcius, R.N., D.B. Eggleston, D.L. Miller, and T.C. Luhrs. 1998. The habitat-survival function for Caribbean spiny lobster: an inverted size effect and non-linearity in mixed algal and seagrass habitats. Marine and Freshwater Research. 49: 807-816.

Maljkovic, A., T.E. Van Leeuwen, and S.N. Cove. 2008. Predation on the invasive red lionfish, Pterois volitans (Pisces: Scorpaenidae), by native groupers in the Bahamas. Coral Reefs. 27: 501.

Marx, J.M. and W.F. Herrnkind. 1986. Species profiles: life histories and environmental requirements of coastal fishes and invertebrates (south Florida) – spiny lobster. U.S. Fish and Wildlife Service Biological Report. 82(11.61). U.S. Army Corps of Engineers, TR EL-82-4. 21 pp.

Maxwell, K.E., T.R. Matthews, R.D. Bertelsen, and C.D. Derby. 2009. Using age to evaluate reproduction in Caribbean spiny lobster, Panulirus argus, in the Florida Keys and Dry Tortugas, United States. New Zealand Journal of Marine and Freshwater Research. 43: 139-149.

McLean, J.H. 1962. Sublittoral ecology of kelp beds of the open coast near Carmel, California. Biological Bulletin. 122: 95-114.

Menge, B.A. 1976. Organization of the New England rocky intertidal community: role of predation, competition, and environmental heterogeneity. Ecological Monographs. 46: 355-393.

Menge, B.A. 1995. Indirect effects in marine rocky intertidal interaction webs: patterns and importance. Ecological Monographs. 65: 21-74.

Menge, B.A. 2000. Top-down and bottom-up community regulation in marine rocky intertidal habitats. Journal of Experimental Marine Biology and Ecology. 250: 257-289.

Miller, B., G. Ceballos, and R. Reading. 1994. The prairie dog and biotic diversity. Conservation Biology. 8: 677-681.

151 Miller, T.E. 1994. Direct and indirect species interactions in an early old-field plant community. The American Naturalist. 143: 1007-1025.

Miller, T.E.X. and V.H.W. Rudolf. 2011. Thinking inside the box: community-level consequences of stage-structured populations. Trends in Ecology and Evolution. 26: 457- 466.

Mintz, J.D., R.N. Lipcius, D.B. Eggleston, and M.S. Seebo. 1994. Survival of juvenile Caribbean spiny lobster: effects of shelter size, geographic location and conspecific abundance. Marine Ecology Progress Series. 112: 255-266.

Moe, M.A., Jr. 1969. Biology of the red grouper (Epinephelus morio Valenciennes) from the eastern Gulf of Mexico. Professional Papers Series, Florida Department of Natural Resources, Marine Science Laboratory, St. Petersburg, FL. 95 pp.

Morris, J.A. and J.L. Akins. 2009. Feeding ecology of invasive lionfish (Pterois volitans) in the Bahamian archipelago. Environmental Biology of Fishes. 86: 389-398.

Mumby, P.J., A.R. Harborne, and D.R. Brumbaugh. 2011. Grouper as a natural biocontrol of invasive lionfish. PLoS ONE. 6: e21510.

Munoz, R.C., C.A. Currin, and P.E. Whitfield. 2011. Diet of invasive lionfish on hard bottom reefs of the Southeast USA: insights from stomach contents and stable isotopes. Marine Ecology Progress Series. 432: 181-193.

Nagkawa, S. and I.C. Cuthill. 2007. Effect size, confidence interval and statistical significance: a practical guide for biologists. Biological Reviews. 82: 591-605.

Naiman, R.J., C.A. Johnston, and J.C. Kelley. 1998. Alteration of North American streams by beaver. BioScience. 38: 753-762.

Nevitt, G., N.D. Pentcheff, K.J. Lohmann, and R.K. Zimmer. 2000. Den selection by the spiny lobster Panulirus argus: testing attraction to conspecific odors in the field. Marine Ecology Progress Series. 203: 225-231.

Okuyama, T. and B.M. Bolker. 2007. On qualitative measures of indirect interactions. Ecology Letters. 10: 264-271.

Orth, R.J., K.L. Heck, and J. van Montfrans. 1984. Faunal communities in seagrass beds: a review of the influence of plant structure and prey characteristics on predator: prey relationships. Estuaries. 7: 339-350.

Paine, R.T. 1966. Food web complexity and species diversity. American Naturalist. 100: 65-75.

Paine, R.T. 1969. A note on trophic complexity and community stability. American Naturalist. 103: 91-93.

152 Paine, R.T. 1992. Food-web analysis through field measurement of per capita interaction strength. Nature. 355: 73-75.

Peckarsky, B.L., P.A. Abrams, D.I. Bolnick, L.M. Dill, J.H. Grabowski, B. Luttberg, J.L. Orrock, S.D. Peacor, E.L. Preisser, O.J. Schmitz, and G.C. Trussell. 2008. Revisiting the classics: considering nonconsumptive effects in textbook examples of predator-prey interactions. Ecology. 89: 2416-2425.

Persson, L., P. Amundsen, A.M. de Roos, A. Klemetsen, R. Knudesn, and R. Primicerio. 2007. Culling prey promotes predator recovery – alternative states in a whole-lake experiment. Science. 316: 1743-1746.

Peterson, C.H. and R. Black. 1994. An experimentalist's challenge: when artifacts of intervention interact with treatments. Marine Ecology Progress Series. 11: 289-297.

Potts, J.C., D. Berrane, and J.A. Morris. 2011. Age and growth of Lionfish from the Western North Atlantic. Proceedings of the 61strd Gulf and Caribbean Fisheries Institute.

Power, M.E., D. Tilman, J.A. Estes, B.A. Menge, W.J. Bond, L.S. Mills, G. Daily, J.C. Castilla, J. Lubchenco, and R.T. Paine. 1996. Challenges in the quest for keystones. BioScience. 46: 609-620.

Pusack, T.J. 2013. Coral-reef fishes: insights into larval dispersal and invasion ecology. Ph.D. dissertation. Oregon State University, Corvallis, OR.

R Core Team. 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. http://www.R-project.org.

Randall, J. E. 1967. Food habits of reef fishes of the West Indies. Studies in Tropical Oceanography. 5: 665-847.

Ratchford, S.G. and D.B. Eggleston. 1998. Size- and scale-dependent chemical attraction contribute to an ontogenetic shift in sociality. Animal Behavior. 56: 1027-1034.

Ratchford, S.G. and D.B. Eggleston. 2000. Temporal shift in the presence of a chemical cue contributes to a diel shift in sociality. Animal Behavior. 59: 793-799.

Reichman, O.J. and E.W. Seabloom. 2002. Ecosystem engineering: a trivialized concept? TRENDS in Ecology and Evolution. 17: 308.

Rhyne, A.L. and J. Lin. 2006. A western Atlantic peppermint shrimp complex: Redescription of Lysmata wurdemanni, description of four new species, and remarks on Lysmata rathbunae (Crustacea: Decapoda: Hipploytidae). Bulletin of Marine Science. 79: 165- 204.

153 Ripple, W.J. and R.L. Beschta. 2012. Trophic cascades in Yellowstone: the first 15 years after wolf reintroduction. Biological Conservation. 145: 205-213.

Rosell, F., O. Bozser, P. Collen, and H. Parker. 2005. Ecological impact of beavers Castor fiber and Castor canadensis and their ability to modify ecosystems. Mammal Review. 35: 248- 276.

Ruttenberg. B.I., P.J. Schofield, J.L. Akins, A. Acosta, M.W. Feeley, J. Blondeau, S.G. Smith, and J.S. Ault. 2012. Rapid invasion of Indo-Pacific lionfishes (Pterois volitans and Pterois miles) in the Florida Keys, USA: evidence from multiple pre- and post-invasion data sets. Bulletin of Marine Science. 88: 1051-1059.

Samhouri, J.F., M.A. Steele, and G.E. Forrester. 2009. Inter-cohort competition drives density dependence and selective mortality in a marine fish. Ecology. 90: 1009-1020.

Schmitz, O.J., V. Krivan, and O. Ovadia. 2004. Trophic cascades: the primacy of trait-mediated indirect interactions. Ecology Letters. 7: 153-163.

Schofield P.J. 2009. Geographic extent and chronology of the invasion of non-native lionfish (Pterois volitans [Linnaeus 1758] and P. miles [Bennett 1828]) in the Western North Atlantic and Caribbean Sea. Aquatic Invasions. 4:473–479.

Schratwieser, J. 1999. The impact of resident and transient predators on the population dynamics of juveniles Caribbean spiny lobster (Panulirus argus) in Florida Bay, Florida. Masters Thesis, Old Dominion University, Norfolk, VA.

Selgrath, J.C., K.A. Hovel, and R.A. Wahle. 2007. Effects of habitat edges on American lobster abundance and survival. Journal of Experimental Marine Biology and Ecology. 353: 253- 264.

Shulman, M.J. 1985. Recruitment of coral reef fishes: effects of distribution of predators and shelter. Ecology. 66: 1056-1066.

Shulman, M.J. and J.C. Ogden. 1987. What controls tropical reef fish populations: recruitment or mortality? An example in the Caribbean reef fish Haemulon flavolineatum. Marine Ecology Progress Series. 39: 233-342.

Simenstad, C.A., J.A. Estes, and K.W. Kenyon. 1978. Aleuts, sea otters, and alternate stable- state communities. Science. 200: 403-411.

Smith, S.G., J.A. Ault, J.A. Bohnsack, D.E. Harper, J. Luo, and D.B. McClellan. 2011. Multispecies survey design for assessing reef-fish stocks, spatially explicit management performance, and ecosystem condition. Fisheries Research. 109: 25-41.

Sogard, S.M. 1997. Size-selective mortality in the juvenile stage of teleost fishes: a review. Bulletin of Marine Science. 60: 1129-1157.

154 Stallings, C.D. 2008. Indirect effects of an exploited predator on recruitment of coral-reef fishes. Ecology. 89: 2090-2095.

Stallings, C.D. 2009a. Predator identity and recruitment of coral-reef fishes: indirect effects of fishing. Marine Ecology Progress Series. 383:251-259.

Stallings, C.D. 2009b. Fishery-independent data reveal negative effect of human population density on Caribbean predatory fish communities. PLoS ONE. 4(5): e5333.

Stier, A.C., S.W. Geange, K.M. Hanson, and B.M. Bolker. 2013. Predator density and timing of arrival affect reef fish community assembly. Ecology. 94:1057–1068.

Steele, M.A. 1997. Regulation by post-settlement mortality in two temperate reef fishes. Oceologia. 112: 64-74.

Travis, J., F.C. Coleman, P.J. Auster, P.M. Cury, J.A. Estes, J. Orensanz, C.H. Peterson, M.E. Power, R.S. Steneck, and J.T. Wootton. 2014. Integrating the invisible fabric of nature into fisheries management. Proceedings of the National Academy of Sciences of the United States of America. 111: 581-584.

Trussell, G.C., P.J. Ewanchuk, and C.M. Matassa. 2006. Habitat effects on the relative importance of trait- and density-mediated indirect interactions. Ecology Letters. 9: 1245- 1252.

Wahle, R.A. and L.S. Incze. 1997. Pre- and post-settlement processes in recruitment of the American lobster. Journal of Experimental Marine Biology and Ecology. 217: 179-207.

Wahle, R.A. 1992. Body-size-dependent anti-predator mechanisms of the American lobster and a model of their trade-offs. Oikos. 65:52-60.

Wahle, R.A. 2003. Revealing stock-recruitment relationships in lobsters and crabs: is experimental ecology the key? Fisheries Research. 65: 3-32.

Weaver, D.C. 1996. Feeding ecology and ecomorphology of three sea basses (pisces: Serranidae) in the Gulf of Mexico. MS thesis. University of Florida, Gainesville, Florida.

Welinder, B.S. 1974. The crustacean cuticle – I. Studies on the composition of the cuticle. Comp. Biochemical Physiology A. 47: 779-787.

Werner, E.E., J.F. Gilliam, D.J. Hall, and G.G. Mittelbach. 1983. An experimental test of the effects of predation risk on habitat use in fish. Ecology. 64: 1540-1548.

Werner, E.E. and J.F. Gilliam. 1984. The ontogenetic niche and species interactions in size- structured populations. Annual Review of Ecology, Evolution, and Systematics. 15:393- 425.

155 Werner, E.E. and S.D. Peacor. 2003. A review of trait-mediated indirect interactions in ecological communities. Ecology. 84: 1083-1100.

White, C.A., C.E. Olmsted, and C.E. Kay. 1998. Aspen, elk, and fire in the Rocky Mountain National Parks of North America. Wildlife Society Bulletin. 26: 449-462.

White, J.S. and J.L. O’Donnell. 2010. Indirect effects of a key ecosystem engineer alter survival and growth of foundation coral species. Ecology. 91: 3538-3548.

White, J.W. 2007. Spatially correlated recruitment of a marine predator and its prey shapes the large-scale pattern of density-dependent prey mortality. Ecology Letters. 10: 1054-1065.

Wilbur, H.M. 1997. Experimental ecology of food webs: complex systems in temporary ponds. Ecology. 78: 2279-2302.

Wilby, A. 2002. Ecosystem engineering: a trivialized concept? TRENDS in Ecology and Evolution. 17: 307.

Wootton, J.T. 1993. Indirect effects and habitat use in an intertidal community: interaction chains and interaction modifications. The American Naturalist. 141: 71-89.

Wootton, J.T. 1994. The nature and consequences of indirect effects in ecological communities. Annual Review of Ecology and Systematics. 25: 443-466.

Wright, J.P., C.G. Jones, and A.S. Flecker. 2002. An ecosystem engineer, the beaver, increases species richness at the landscape scale. Oceologia. 132: 96-101.

Wright, J.P. and C.G. Jones. 2004. Predicting effects of ecosystem engineers on patch-scale species richness from primary productivity. Ecology. 85: 2071-2081.

Wright, J.P. 2009. Linking populations to landscapes: richness scenarios resulting from changes in the dynamics of an ecosystem engineer. Ecology. 90: 3418-3429.

Valdez-Moreno, M., C. Quintal-Lizama, R. Gomez-Lozano, and M.C. Garcia-Rivas. 2012. Monitoring an alien invasion: DNA barcoding and the identification of lionfish and their prey on coral reefs of the Mexican Caribbean. PLoS ONE. 7(6): e36636.

Venables, W.N. and B.D. Ripley. 2002. Modern Applied Statistics with S. Fourth Edition. Springer, New York. ISBN 0-387-95457-0.

Zieman, J.C., J.W. Fourqurean, and R.L. Iverson. 1989. Distribution, abundance and productivity of seagrasses and macroalgae in Florida Bay. Bulletin of Marine Science. 44:292–311.

156 BIOGRAPHICAL SKETCH

Robert Dodge Ellis was born in Honolulu, Hawaii, in April 1981. He in Long Beach,

California, where he graduated from Long Beach Polytechnic High School in June 1999. He attended the University of California Santa Barbara from 1999 to 2004, and spent his final

undergraduate year as an Education Abroad Program participant at the University of Western

Australia in Perth, Australia. In March 2004 he received the degree of Bachelor of Science in

Aquatic Biology and was awarded Distinction in the Major for the completion of an honor’s thesis on the site fidelity and habitat associations of small-bodied groupers in St. Croix, USVI.

After three years working in experiential and K-12 education and after a brief foray in international hospitality, he entered the Master’s program at Louisiana State University in

August 2007. In June of 2009 he successfully defended a Master’s thesis that focused on the population dynamics of a protogynous grouper and efficacy of spatial closures for fisheries management. He was awarded the degree of Master of Science in Oceanography in August 2009 after which he immediately entered the Doctoral program at the Florida State University. From

2010 through 2014 he conducted research on the ecology of groupers in south Florida. In

February 2015 Robert accepted a John A. Knauss Marine Policy Fellowship with the National

Marine Fisheries Service in Silver Spring, MD. In the fall of 2015 he successfully defended his dissertation. Following completion of his fellowship in 2016, Robert plans to pursue a career as a research scientist.

157