Foundation Species as Drivers of Ecosystem Structure, Multifunctionality, and Resilience

Material Information

Foundation Species as Drivers of Ecosystem Structure, Multifunctionality, and Resilience
Angelini, Christine
Place of Publication:
[Gainesville, Fla.]
University of Florida
Publication Date:
Physical Description:
1 online resource (162 p.)

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Committee Chair:
Committee Co-Chair:
Committee Members:
Graduation Date:


Subjects / Keywords:
Aggregation ( jstor )
Crabs ( jstor )
Drought ( jstor )
Ecology ( jstor )
Ecosystems ( jstor )
Invertebrates ( jstor )
Mud flats ( jstor )
Mussels ( jstor )
Salt marshes ( jstor )
Species ( jstor )
Biology -- Dissertations, Academic -- UF
biodiversity -- crab -- epiphyte -- facilitation -- function -- invertebrate -- marsh -- mussel -- spartina -- trophic
City of Live Oak ( local )
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Zoology thesis, Ph.D.


Foundation species are dominant, structure-forming organisms that modify physical and biotic conditions to facilitate associated communities { TC ABSTRACT }. Although foundation species nearly always co-occur, our understanding of how they interact with each other and collectively influence how ecosystems are structured, function, and respond to disturbance remains limited. Here, I synthesize research on foundation species interactions and quantify the effect that facilitation cascades, interaction chains in which primary foundation species facilitate secondary foundation species, have on ecosystem structure, functioning, and resilience. Many studies have either examined how competition and facilitative interactions drive spatial segregation and overlap, respectively, of foundation species, or what effects a single foundation species or functional group have on associated community characteristics. Integrating these complementary fields of research, I propose that the nature of interactions among foundation species controls landscape patterns in community structure. I then present results from experiments and surveys that assess mechanisms of facilitation that structure a tree (Quercus virginiana)- epiphyte (Tillandsia usneoides) cascade and examine its effect on invertebrate species, functional group, and life stage diversity. I discovered that Tillandsia relies on physical stress reduction provided by Quercus, and secondarily reduces multiple stressors to enhance all aspects of arboreal invertebrate diversity. To assess whether the presence of multiple foundation species may also regulate ecosystem functioning, I tested how the density of ribbed mussels (Geukensia demissa) that occur in aggregations within salt marsh habitat formed by cordgrass (Spartina alterniflora) influences salt marsh invertebrate diversity, 7 ecosystem functions, and multifunctionality, the simultaneous performance of these functions. Using surveys of mussel distribution, I then extrapolated my experimental results to estimate the net effect of mussels, and the diverse invertebrate communities they facilitate, on multifunctionality at the landscape scale. In my final study, I used monitoring, model simulations, and experiments to demonstrate that the resilience of southeastern US salt marshes to drought hinges on the presence and spatial distribution of remnant cordgrass patches and that mussels, in facilitating cordgrass survival during drought, are likely playing a keystone role in regulating the resilience of this ecosystem. Combined, my research advances our understanding of how the interactions among foundation species regulates biodiversity, multifunctionality, and resilience within natural ecosystems. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis (Ph.D.)--University of Florida, 2014.
Co-adviser: PALMER,TODD.
Electronic Access:
Statement of Responsibility:
by Christine Angelini.

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
Embargo Date:
Resource Identifier:
968786006 ( OCLC )
LD1780 2014 ( lcc )


This item has the following downloads:

Full Text




© 2014 Christine Angelini


To my many mentors


4 ACKNOWLEDGMENTS This work has been inspired, propelled, and shaped by diverse group of mentors. My advisor, Brian Silliman, has provided me with outstanding opportunities to grow intellectually and professionally , a loving lab community , and continued encouragement. Under his guidance, I have learned to simplify and strengthen my id eas and better communicate my science. Brian has inspired me as an ecologist, teacher, parent, and friend and I am so grateful for the experience I have had as a member of his lab . I have also received valuable guidance and support from my committee member s: Luke Flory, Doug Levey, Craig Osenberg , and Todd Palmer . As a result of their feedback, t he quality of the work that I have produced during this time has been significantly enhanced and my skills as an ecologist, writer, and colleague sharpened . I am pa rticularly appreciative of Craig for challenging me to think critically and clearly from my very first day as a grad student and including me in hi s lab group ; our interactions have been an integral, special part of my experience in this department . I am also grateful for the camaraderie and intellectual contributions of Marc Hensel, Tjisse van der Heide, John Griffin, Leon Lamers, Michael McCoy, Liz Schrack, James Nifong, Hannah van der Zanden, Alfons Smolders, and Johan van de Koppel. These friends and c olleagues have not only made conducting this research incredibly fun, but also cultivated my technical skills and greatly expanded my knowledge . I a m also a ppreciative of the effort and enthusiasm of those who worked with me in the field and lab : Eric Mona co, Robbie McNulty, Nicole Soomdat, Jackie Babb, Emma Knight, Rebecca Atkins, Kristin Briggs, Tessa Diehl, Alissa Mazzoli, Corrinne Fuchs, Jessica Mulvey, Steph anie Buhler, Hannah Nelson , Brian Gibbs, Helene de Paoli, and Marlous Hooyiberg. With the sweat and friendship of these students, I have been able to


5 accomplish far more than I ever expected when I began this journey . I also appreciate the talent and time provided by G.B Edwards, Gary Ste ck, and John Slapcinsky, whose impressive expertise allowed me to more thoroughly characterize the Tillandsia arthropod communities . In addition, I thank Fred Diehl, Bev Marcum, Bob McNulty, and Maggie Hunter for sharing their passion for educatio n and the natural world with me . Our times together in San Sal have been some of the most positive and influential experiences of my life. The l ogistical support provided by Gracie Townsend , Richard Alston , Jason Johnson , Ike Sellers , Jacob Shalack , Dorset Hurley , and Caroline Reddy on Sapelo, and Susan Spaulding , Tangelyn Mit chell , and Quintina Meekins at UF has been invaluable as well . I also thank the Sapelo Island National Estuarine Research Reserve, the Georgia Coastal Ecosystems LTER, Francis Marion National Forest, Ron and Rita Phillips , and Dale Aren for granting me acc ess to field sites and permission to move aroun d a lot of marsh , moss , and trees . Funding for my dissertation research has been provided by the University of Florida Graduate Alumni Fellowship , UF Graduate Student Study Abroad Travel Grant, NSF GRFP fellow ship (DGE 0802270) , and NSF Career Award and Biological Oceanography funding awarded to Brian Silliman . Finally, I thank my husband, Tommy Angelini, parents, David and Macyln, siblings, Bert, Joanna, Russ, Geoff and Evan, extended family, Ann, Brian, Wes, Kate, Kim, Matt, and RT, and close friends, Rebecca and Haven, for their love , patience, and support.


6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 Theoretical Context ................................ ................................ ................................ . 15 Purpose ................................ ................................ ................................ .................. 17 Outline ................................ ................................ ................................ .................... 17 2 INTERACTIONS AMONG FOUNDATION SPECIES AND THE IR CONSEQUENCES FOR COMMUNITY ORGANIZATION, BIODIVERSITY AND CONSERVATION ................................ ................................ ................................ ... 20 Introduction to Foundation Species ................................ ................................ ........ 20 Coexistence of Foundation Species ................................ ................................ ....... 24 Strength and Direction of Foundation Species Interactions ................................ .... 26 Facilitation ................................ ................................ ................................ ........ 26 Competition ................................ ................................ ................................ ...... 29 Effects of Foundation Species Diversity and Habitat Complexity on Biodiversity ... 30 T he Ecological Theater and Evolutionary Play: the Context Dependence of Foundation Species Interactions ................................ ................................ ......... 31 Case example 1 ................................ ................................ ......................... 31 Case e xample 2 ................................ ................................ ......................... 32 Conclusions ................................ ................................ ................................ ............ 33 3 SECONDARY FOUNDATION SPECIES AS DRIVERS OF TROPHIC AND FUNCTIONAL DIVERSITY: EVIDENCE FROM A TREE EP IPHYTE SYSTEM .... 37 Introduction ................................ ................................ ................................ ............. 37 Methods ................................ ................................ ................................ .................. 40 Oak Facilitation of Til landsia : an Experiment ................................ .................... 40 Tillandsia Facilitation of Invertebrates: the Physical Environment .................... 42 Tillandsia Facilitation of Inverte brates: Survival ................................ ................ 43 Oak Tillandsia Facilitation of Invertebrates: Community Responses .............. 44 Generality of Oak Tillandsia Facilita tion of Invertebrates: :Latitudinal Survey ................................ ................................ ................................ ........... 45 Results ................................ ................................ ................................ .................... 46


7 Discussion ................................ ................................ ................................ .............. 48 Mechanisms Maintaining Tree Epiphyte Facilitation Cascades ....................... 49 Secondary Foundation Species as Drivers of Trophic Structure and Ecosystem Functioning ................................ ................................ ................. 52 Foundation Species Biodiversity (FSB) Model ................................ ................. 53 4 SECONDARY FOUNDAITON SPECIES GENERATE HOTSPOTS OF MULTIFUNCTIONALITY IN A COASTAL ECOSYSTEM ................................ ........ 67 Introduction ................................ ................................ ................................ ............. 67 Methods ................................ ................................ ................................ .................. 70 Study System ................................ ................................ ................................ ... 70 Density Dependent Effects of Mussels on Biodiversity and Multiple Ecosystem Functions: an Experiment ................................ ........................... 71 Invertebrate functional group diversity ................................ ....................... 72 Invertebrate biomass ................................ ................................ ................. 73 Soil accretion ................................ ................................ ............................. 73 Decomposition ................................ ................................ ........................... 74 Infiltration rate ................................ ................................ ............................ 74 Benthic algae biomass ................................ ................................ ............... 74 Aboveground cordgrass biomass ................................ ............................... 75 Belowground cordgrass biomass ................................ ............................... 75 Multifunctionality ................................ ................................ ........................ 75 Analyses ................................ ................................ ................................ ........... 76 Effect of Mussels on Biodiversity, in turn, on Multifunctionality ......................... 77 Landscape Level Effects of Mussel Aggregations on Biodiversity and Multifunctionality: a Survey ................................ ................................ ............ 78 Results ................................ ................................ ................................ .................... 78 Discussion ................................ ................................ ................................ .............. 80 Density Dependent Effects of Mussels on Invertebrate Diversity and Ecosystem Functions ................................ ................................ .................... 81 The Importance of Mussels as Drivers of Diversity and Multifunctionality at Landscape Scales ................................ ................................ ......................... 84 Integrating Secondary Foundation Species into Natural Resource Management ................................ ................................ ................................ . 87 5 REMNANT PATCHES, A KEYSTONE MUTUALISM, AND THE RESILIENCE OF A COASTAL ECOSYSTEM TO DRO UGHT ................................ ..................... 93 Introduction ................................ ................................ ................................ ............. 93 Methods ................................ ................................ ................................ .................. 96 Study System: Southeastern US Sa lt Marshes ................................ ................ 96 Remnant Patch Effects on Mudflat to Cordgrass Transitions: the Salt Marsh Recovery Model ................................ ................................ ............................ 97 Regional Survey of Cordg rass Mortality and Survival: are Mussels Enhancing Patch Resistance? ................................ ................................ .... 100 Mechanisms by which Mussels Enhance Cordgrass Resistance to Drought . 102


8 The Contribution of Mussel Mutualists in Driving Mudflat Recovery ............... 104 Results ................................ ................................ ................................ .................. 105 Discussion ................................ ................................ ................................ ............ 107 Projecting Salt Marsh Recovery from Drought: are Remnant Patches Important? ................................ ................................ ................................ ... 108 Cordgrass Resistance to Drought: Extent of Mortality and the Mechani sms of Survival ................................ ................................ ................................ ... 109 Effect of Mussel Mutualists on Cordgrass Recovery at Patch and Mudflat Scales ................................ ................................ ................................ ......... 112 Conclusions ................................ ................................ ................................ .......... 113 APPENDIX A MECHANISMS OF CORDGRASS FACILITATION OF MUSSELS ...................... 121 B RELATIONSHIP BETWEEN MUSSEL AGGREGATION AREA AND MUSSEL DENSITY ................................ ................................ ................................ .............. 123 C SOIL DEPOSITION MEASUREMENT ................................ ................................ .. 124 D SOIL AMMONIA AND ORGANIC CARBON MEASUREMENTS .......................... 125 E SUMMARY OF NULL, LINEAR, LOG, HYPERBOLIC, AND POWER MODEL COMPARISONS ................................ ................................ ................................ ... 127 F FREQUENCY OF MUSSEL AGGREGATIONS OF DIFFERENT SIZES AT TWO SAPELO ISLAND SALT MARSH PLATFORMS ................................ .......... 131 G CORDGRASS RECOVERY FROM SEEDS AND CLONAL GROWTH: AN EXPERIMENT ................................ ................................ ................................ ...... 132 H SAPELO ISLAND MUDFLAT CHARACTERISTICS ................................ ............. 135 I SPATIAL DISTRIBUTION OF PATCHES IN SAPELO ISLAND MUDFLATS ....... 136 J EFFECT OF MUSSELS ON CORDGRASS RECOVERY AT THE PATCH SCALE ................................ ................................ ................................ .................. 137 Methods ................................ ................................ ................................ ................ 137 Results ................................ ................................ ................................ .................. 138 Implications ................................ ................................ ................................ ........... 140 K R CODE FOR THE SPARTINA RECOVERY MODEL ................................ ......... 141 L SUMMARY OF DROUGHT GENERATED MUDFLATS AND REMNANT PATCH COVER ACROSS THE SOUTHEASTERN US COAST .......................... 146 M RAINFALL IN SUMMER 2012: SALINITY MONITORING PERIOD ..................... 147


9 LIST OF REFERENCES ................................ ................................ ............................. 148 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 162


10 LIST OF TABLES Table page 3 1 Invertebrate response to Tillandsia . ................................ ................................ .... 57 3 2 Effects of Tillandsia acro ss latitudes. . . ................................ ................................ 58 3 3 Results from latitudin al survey rarefaction analyses. . ................................ ......... 58 4 1 Effects of mussels on biodive rsity and ecosystem functions . ............................. 89 E 1 Model comparisons: m ussels density effects on biodiveristy . ........................... 127 E 2 Model comparisons: musse l density effects on individual ecosystem functions . ................................ ................................ ................................ .......... 128 E 3 Model comp arisons: mussel density effects on multifunctionality indices ......... 129 E 4 Best fit models:m ussel density effects on mutlifuncitonality threshold indices . . 130 E 5 Model comparisons: invertebrate d iversity and multifunctionality indices . ........ 130 H 1 Characteristics of 9 Sapelo Island, GA mudflats.. ................................ ............. 135 L 1 Summary of c ordgrass die off and resistance across SE US coast. . ................ 146


11 LIST OF FIGURES Figure page 2 1 Types of foundation species assemblages ................................ ......................... 35 2 2 Context dependence o f interactions among foundation species . ........................ 36 3 1 Live oak draped with Spanish moss ................................ ................................ ... 59 3 2 Method for scoring Tillandsia transplant survival ................................ ................ 60 3 3 Live oak and experimental shade mimic effects on light ................................ ..... 61 3 4 Mechanisms of oak facilitation of Tillandsia ................................ ........................ 62 3 5 Mechanisms of oak facilitation of Tillandsia ................................ ........................ 63 3 6 Tillandsia effects on canopy temperature and humidity . ................................ .... 64 3 7 Oak Tillandsia facilitation cascades across latitude ................................ ............ 65 3 8 The Foundation Species Bio diversity model ndary foundation ........................... 66 4 1 Effect of mussel addition on invertebrate diversity and multifunctionality . .......... 90 4 2 Effect of mussels on diversity and diversity on multifunctionality . ....................... 91 4 3 Landscape effects of mussels on marsh diversity and multifunctionality ............ 92 5 1 Remnant patch dynamics ................................ ................................ ................. 115 5 2 Drought trends and cordgrass die off ................................ ............................... 116 5 3 Sa lt Marsh Recovery model results ................................ ................................ .. 117 5 4 Coastal survey of remnant patch association with mussels .............................. 118 5 5 Relationship between mussels and salinity ................................ ...................... 119 5 6 Contrib ution of mussel associated patches to mudflat recovery . ...................... 120 A 1 Mechanisms of cordgrass facilitation of mussels.. ................................ ............ 122 B 1 Relation ship between aggregation area and the number of mussels. .............. 123 C 1 E ffect of mussel density on soil deposition rate. ................................ ............... 124 D 1 Effect of mussel density on soil organic C and porewater ammonia ................. 126


12 E 1 Effect of invertebrate diversity on multifunctionality threshold indices . ............. 130 F 1 Mussel distribution survey in two Sapelo Island, GA marsh platforms. ............. 131 G 1 Cordgrass recovery from seeds and clonal ramets . ................................ ......... 134 I 1 Summary of the remnant patch distribution in 4 mudflats.. ............................... 136 J 1 Effect of mussels on patch expansion. ................................ ............................. 139 M 1 Daily precipitation on Sapelo Island, GA over summer 2012 ............................ 147


13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FOUNDATION SPECIES AS DRIVERS OF ECOSYSTEM STRUCTURE, MULTIFUNCTIONALITY, AND RESILIENCE By Christine Angelini August 2014 Chair: Brian Reed Silliman Major: Zoology Foundation species are dominant, structure-forming organisms that modify physical and biotic conditions to facilitate associated communities . Although foundation species nearly always co-occur, our understanding of how the y interact with each other and collectively influence how ecosystems are structured, function, and respond to disturbance remains limited. Here, I synthesize research on foundation species interactions and quantify the effect that facilitation cascades, interaction chains in which primary foundation species facilitate secondary foundation species, have on ecosystem structure, functioning, and resilience. Many studies have either examined how competition and facilitative interactions drive spatial segregation and overlap, respectively, of foundation species, or what effects a single foundation species or functional group have on associated community characteristics . Integrating these complementary fields of research, I propose that the nature of interactions among foundation species controls landscape patterns in community structure. I then present results from experiments and surveys that assess mechanisms of facilitation that structure a tree (Quercus virginiana )epiphyte ( Tillandsia usneoides) cascade and


14 examine its effect on invertebrate species, functional group , and life stage diversity. I discovered that Tillandsia relies on physical stress reduction provided by Quercus , and secondarily reduces multiple stressors to enhance all aspects of arboreal invertebrate diversity . To asses s whether the presence of multiple foundation sp ecies may also regulate e cosystem function ing, I tested how the density of ribbed mussels ( Geukensia demissa ) that occur in aggregations within salt marsh habitat formed by cordgrass ( Spartina alterniflora ) influences salt marsh invertebrate diversity , 7 e cosystem functions, and multifunctionality, the simultaneous performance of these functions. Using surveys of mussel distribution, I then extrapolated my experimental results to estimate the net effect of mussels, and the diverse invertebrate communities t hey facilitate, on multifunctionality at the landscape scale . In my final study, I used m onitoring, model simulations, and experiments to demonstrate that the resilience of southeastern US salt marshes to drought hinges on the presence and spatial distribu tion of remnant cordgrass patches and that mussels, in facilitating cordgrass survival during drought , are likely playing a keystone role in regulating the resilience of this ecosystem . C ombined , my research advances our understanding of how the interacti ons among foundation species regulates biodiversity, multi functionality, and resilience within natural ecosy s tems.


15 CHAPTER 1 INTRODUCTION Theoretical Context Positive interactions , including facultative and obligatory facilitation and mutualisms, are no n consumer interactions among two or more species that positively affect at least one of the species involved (Bertness and Callaway 1994) . Amelioration of physical and consumer stress , pollination , and nutritional resource exchanges, such as those between plants and mycorrhizal fungi, are common types of positive interactions , all of whi ch succession, and distributions (Ellison et al. 2005, Br ooker et al. 2008, He et al. 2013) . Due to the prevalence of positiv e interactions and their important role in controlling the structure and organization of many natural communities, there have been concerted efforts to incorporate them into ecological th eory (Bruno et al. 2003) and, more recently, restoration and conservation (Halpern et al. 2007) . One prominent manifestation of positive interactions is that of foundation species (Dayton 1972) , habitat forming organisms that alleviate physical stress and consumer p ressure to facilitate associated communities . Most research conducted to date on this topic has investigated the eff ects of a single foundation species or functional group on community structure. A wide variety of ecosystems are characterized by multiple foundation species but have not traditionally been described or managed as such because of the scale at which ecologi sts typically conduct experiments, inconsistent use of terms associated with foundation species, and infrequent application of this concept in studies outside of marine ecology (Ellison et al. 2005). The spatial scale of experiments often leads to the over sight of multiple foundation species effects, b ecause


16 foundation species are distributed widely across ecosystems, as in mangrove forests and salt marshes, in which multiple species are arranged in broad zones of dominance and experiments are done on small er spatial scales within zones, or because experiments are conducted entirely within a community defined by a foundation species assemblage and the interactive effects of habitat modifying foundation species are not considered (Altieri et al. 2007). Termin ology problems have also diluted the focus on foundation species effects. Dayton (1972) coined the term foundation species nearly four decades ago, but widespread use of similar terms, such as dominant species (Grime 1987) and ecosystem engineer (Jones et al. 1997) has blurred the concept. In addition to being both dominant species (productive organisms that garner a disproportionate share of resources and competitively exclude subordinate species) and autogenic ecosystem engineers (organisms that change ab iotic and biotic conditions through their own physical structure), foundation species have strong, positive effects on many other organisms in the community. I use the term foundation species because it has historical precedence and identifies a class of o rganism without which the associated biological community would not persist. In addition, many ecologically important, structure generating organisms are often not recognized as foundation species. Because I have defined foundation species as organisms tha t provide structure; moderate local biotic and abiotic conditions; and have a large, positive effect on other species in a community (Dayton 1972), clams that provide a hard substrate for the attachment of sessile invertebrates within soft sediment habitat s (Gribben et al. 2009) and arboreal epiphytes, Asplenium nidus ), which harbor diverse invertebrates from


17 predation and physical stress (Ellwood and Foster 2004), are considered foundation species. As a result of these types of o versight, the prevalence and importance of assemblages of multiple foundation species has been vastly underestimated. Purpose Using field and mesocosm experiments, surveys, and models, I examine the effects of foundation species and their interactions on e cosystem structure, functioning, and resilience in two model communities: southern live oaks ( Quercus virginiana ) laden with Spanish moss ( Tillandsia usneiodes ) in coastal savannas and salt marshes in the southeastern United States dominated by smooth cord grass ( Spartina alterniflora ). Outline In Chapter 1, I provide a review of the literature on what is known about interacti ons among foundation species and explore how the nature of interactions among foundation species that co occur in an ecosystem may con trol spatial patterns in the distribution of biodiversity. In particular, I hypothesize that competition among foundation species leads to segregation in their distribution as well as the communities of organisms each foundation species supports, while fac ilitation leads to nested assemblages of foundation species that support communities that are far more diverse than those supported by a single, dominant foundation species alone. To conclude, I explore how conservation and management strategies may be imp roved by taking a multiple foundation speci In Chapter 2, I u se experiments and a survey of six sites distributed from Florida to North Carolina to examine the effects of facilitation among foundation species on species and trophic diversity. Specifically, I test if and how southern live oaks facilitate Spanish moss and Spanish moss, in turn, facilitates invertebrate communities. I


18 discovered that live oaks facilitate Spanish moss by alleviating otherwise lethal temperature and light stress, an d Spanish moss facilitates invertebrates by alleviating consumer pressure and humidity stress. Invertebrate species abundance and trophic diversity within oaks was strongly enhanced by the presence of Spanish moss in both experiments and surveys, sugg estin g that this tree (Altieri et al. 2007) has powerful effects on controlling biodiversity in coastal savannas . In Chapter 3, I explore context dependence in the effects of foundation species on both biodiversity and multiple ecosystem functions. Specifically, I manipulated the number of ribbed mussels that form aggregations in a cordgrass monocul ture and to measure the density dependent effects of this secondary foundation species (i.e. one that is obligately dependent on a primary foundation species) on biodiversity, seven salt marsh ecosystem functions and the simultaneous performan ce of all of these functions multifunctionality. Using surveys of mussel distribution, I then extrapolate my experimental results to explore how mussels, at their natural distribution, may control salt marsh biodiversity and ecosystem functioning at the landscape scal e. Inspired by my observation that cordgrass survived in remnant patches within mudflats generated by a severe drought , and did so with particularly high frequency when it was associated with mussels, I then used surveys, correlational and experimental ap proaches, and simulation models to test if remnant patches are accelerating salt marsh recovery and if mussels are functioning as key mutualists that permit cordgrass to survive drought. Using field collected data to parameterize a Salt Marsh Recovery mode l, I found that the size of mudflats as well as the spatial distribution of cordgrass patches surviving within those mudflats interact to control how


19 long it will take for mudflats to recover to cordgrass dominance. In addition, monitoring and experiments revealed that mussels are likely enhancing cordgrass resistance to drought by simultaneous enhancing the availability of their growth limiting nutrient (Nitrogen) and alleviation of a major abiotic stress for cordgrass (i.e. dry soils) . Although mussel mu tualists do not stimulate cordgrass recolonization at scale of individual remnant patches, through their support of many, spatially dispersed patches, they are likely substantially augmenting to the ability of salt marshes to recover from massive , drought induced disturbance. Collectively, this body of work provides general insight to the importance of positive interactions, particularly those that involve foundation species, in mediating patterns in biodiversity, ecosystem functioning, and ecosystem resili en ce to intensive climate stress.


20 CHAPTER 2 INTERACTIONS AMONG FOUNDATION SPECIES AND THEIR CONSEQUENCES FOR COMMUNITY ORGANIZATION, BIODIVERSITY AND CONSERVATION Introduction to Foundation Species Ecologists have long recognized the role of foundation species in facilitating whole communities of organisms through habitat creation (Dayton 1972, Bertness and Callaway 1994, Stachowicz 2001, Ellison et al. 2005) . Kelps, conifers, and corals, for instance, are spatially dominant organisms whose biogenic structure promotes species coexistence through the amelioration of physi cal stress and the creation of fine scale, complex matrices where smaller organisms find refuge from predators and competitors (Dayton 1972, Stachowicz 2001) . Foundation species are often primary producers or bed forming fi lter feeders and play central roles in sustaining ecosystem services, such as nursery habitat for fish (Boesch and Turner 1984, Carr 1989, Beck et al. 2001), shoreline stabilization (Orth et al. 2006, Koch et al. 2009) , water filtration (Altieri and Witman 2006) , timber production and carbon sequestration (Ellison et al. 2005) . Due to the economic value of these services and the link between foundation species and biodiversity, recent conservation strategies have prioritized protecting and restoring foundation species in degraded ecosystems (Crain and Bertness 2006, Byers et al. 2006, Halpern et al. 2007, Gómez Aparicio 2009) . To date, most of our understanding of how foundation species affect community organization and biodiversity has emerged from studies of marine ecosystems that investigate facu ltative effects of a single dominant space holder. In coastal soft sediment habitats, for instance, extensive beds of the seagrass, Thalassia testudinum, can cover otherwise low productivity, sand flats occupied by algae. By reducing flow, modifying substr ate and impeding the foraging efficiency of mobile predators, seagrass


21 beds facilitate the settlement of benthic invertebrates and enhance the survivorship and density of prey species (Orth et al. 1984, Heck et al. 2003, Canion and Heck 2009) . Similarly, along 1000s of kilometers of Caribbean coastline, networks of red mangrove ( Rhizophora mangle ) roots provide essential nursery habitat for juvenile snappers, grunts, barracuda and other fishes, physically guarding these populations from larger predators (Beck et al. 2001, Faunce and Serafy 2008) . Consequently, the trophic dynamics associated with seagrass and mangrove communities largely arise from the biogenic framework provided by foundation species, suggesting hie rarchical community organization in which species and their interspecific interactions occur within a community that is itself established by the facilitation of a foundation species (Bruno and Bertness 2001) . Field experiments that manipulate the presence or mimic the physical attributes of foundation species have elucidated the mechanisms by which foundation species influence the distribution of associated organisms, e.g. shadi ng by nurse plants (Franco and Nobel 1989) , substrate stabilization by cordgrass (Altieri et al. 2007) and nursery effects of mangroves (Laegdsgaard and Johnson 2001) . These studies have motivated the revision of general models of community organization that p reviously emphasized predation, disturbance and competition (Levin and Paine 1974, Menge and S utherland 1987) to include positive (i.e., facilitative) interactions. In particular, recognition of the pervasive role of facilitation in communities has given rise to the stress gradient hypothesis (Bertness and Callaway 1994) , modified predictions made by the fundamental niche, intermediate disturbance and diversity invasion hypothese s (Bruno et al. 2003) and inspired comprehensive reviews of facilitative interactions in a wide range of ecosystems (Callaway 1995, Stachowicz 2001, Maestre et al. 2009) .


22 While fiel d studies, models and syntheses have improved our understanding of the central role that foundation species play in structuring communities, our current approach of examining a given foundation species in isolation or by lumping multiple foundation species into a single functional entity overlooks a key characteristic of community organization. Specifically, most ecosystems are structured by multiple foundation species whose differences in structural and functional morphology influence their community impac t (Bruno and Bertness 2001) . Seagrass meadows are frequently mixed stands of Thalassia, Zostera and/or Enhalus species which vary in structural characteristics and functional tr aits (Duarte 2000) ; coral reefs are composed of multiple encrusting and branching clonal organisms whose growth forms operate in concert to form complex biogenic reef structures and promote species diversity (Knowlton and Jackson 2001) ; Costa Rican cloud forests intermix palm to bam boo dominated communities with increasing altitude with cascading effects on associated flora and fauna ( Kappelle et al. 1995) ; and mixed stands of Australian kelp are more common and harbor more diverse benthic assemblages than monospecific patches (Irving et al. 2004, Goodsell et al. 2004) . Despite the prevalence of multiple foundation species and importance of species specific traits in modifying habitats, few studies h ave evaluated the effects of foundation species assemblages on habitat complexity or the spatial distribution, composition, and persistence of higher trophic levels . In one of the few studies to explicitly evaluate the coexistence of foundation species, Al tieri et al. (2007) examined the interactions between foundation species and tested whether they had additive or redundant roles in facilitating New England cobble beach communities. They found that the foundation species, cordgrass ( Spartina


23 alterniflora ) , could independently colonize the shore and facilitate the establishment of ribbed mussel beds ( Geukensia demissa ) within its biogenic matrix by stabilizing and shading the substrate. Established mussels further buffered evaporative stress and generated h ard substrate, resulting in higher abundance of species that depend on rigid, stable surfaces (e.g. algae, barnacles, blue mussels) relative to cobble areas without cordgrass and mussels (Altieri et al. 2007) . In addition to the abundance of associated organisms, the diversity and overall stability of the cobble beach community is maintained via facilitation cascades in whic h an independent, stress tolerant foundation species (cordgrass) facilitates a second, dependent foundation species (ribbed mussels) to provide complementary levels of complexity (i.e. small and large crevices, hard and soft substrate) and enhance stress a melioration (Altieri and Wesenbeeck 2010) . The regulari ty with which foundation species distributions overlap suggests that emergent effects, like facilitation cascades, may play a critical role in the organization and stabilization of many communities (Yakovis et al. 2008a) . Coexisting foundation species also compete for space and limiting resources. Although many studies have demonstrated competition between dominant, habitat forming space holders (e.g. tropical forests, coral reefs, salt marshes), few have quantified or inferred how competitive interactions among foundation species influence habitat complexity, diversity and community org anization . In this paper we examine: Under what conditions do foundation species coexist? What mediates the strength and direction of foundati on species interactions? And, d oes variation in the nature and strength of foundation species interactions generat e predictable landscape scale patterns in habitat complexity and the distribution of associated species? We use the


24 stress gradient hypothesis (Bertness and Callaway 1994) as a conceptual framework to explore these questions because it has been well supported by studies conducted in wide range of ecosystems (Bruno et al. 2003, He et al. 2013) . We conclude by discussing how a multiple foundation species perspec tive could enhance the success of future conservation efforts. Coexistence of Foundation Species As space holders that generate and modify habitats, foundation species dominate available substrate in most environments and can coexist at stable population d ensities in either nested or adjacent assemblages (Fig. 1). Nested foundation spe cies assemblages occur when: the first foundation species to colonize a habitat does not monopolize the substrate, enabling colonization of a second foundation species in inte rstitial space [i.e., saguaro cacti within a nurse shrub matrix (Turner et al. 1966) , mussels within seagrass (Valentine and Heck 1993) , clams within macroalgal beds (Gribben et al. 2009) ] or the first foundation species to colonize a habitat provides novel substrate for colonization and survival of other foundation species [i.e., tree limbs that host arboreal bromeliads and ferns in neotropical forests (Matelson et al. 1993) , sponges that bind and stabilize rubble to mediate coral attachment and reef growth (Wulff 1984) , and large, foliose brown seaweeds that host a diversity of epiphytic algae (Hay 1986) ]. In both types of nested assemblages, initial modification provided by the first foundation species to colonize a habitat allows for the settlement and success of foundation species that would not otherwise occur under ambient environm ental conditions. Once established, the magnitude and form of habitat modification (e.g., predator refuge, moisture retention, light regulation) provided by coexisting foundation species is typically complementary and differs as a function of species speci fic traits


25 (Irving and Bertness 2009) . For instance, paloverde trees buffer evaporative stress on the obligate foundation species, saguaro cacti, in the Sonoran Desert (Turner et al. 1966) , but contribute far less than cacti as a water and nutrient resource or predation refuge for asso ciated birds and invertebrates (Wolf and Martinez del Rio 2003) . In contrast to classical facilitative interactions where one organism directly enhances the success of another, facilitation among foundation species drives whole community development by generating habi tats with multiple levels of structure and a diversity of resources. In other cases, foundation species assemblages coexist at large scales across adjacent habitats. Adjacent foundation species occ ur when a foundation species monopolizes are as of primary s ubstrate, and inhibits colonization by other foundation species that are unable to utilize the interstitial space or novel substrate created by th e dominant foundation species. These adjacent assemblages are most apparent in patterns of foundation species zonation that occur when ecosystems are viewed at the landscape scale, such as hardwood and conifer zones on mountainsides (Kappelle et al. 1995, Hsieh et al. 2009) , and red, black, and white mangroves that border one another on tr opical coasts (Sousa and Kennedy 2007) . Within each zone, the competitively dominant foundation species locally mediates the complexity of the habitat and drives variation in structural attributes, such canopy height, crevice size or substrate conditions, and community composition across space. As in n ested assemblages, the diversity and abundance of associated organisms are also promoted at the landscape level where multiple, adjacent foundation species persist.


26 Strength and Direction of Foundation Species Interactions Although foundation species are distinguished as special class of organism (Dayton 1972) , they are ultimately limited by physical and/or biological (i.e., competition and predation) stresses or disturbance like any other species (Levin and Paine 1974, Menge and Sutherland 1987) . Consequently, we anticipate that patterns in the strength and direction of foundation species interactions mirror classic ecological interactions among species that are n ot foundation species. The SG H (Bertness and Callaway 1994) predicts that species interactions are negative (compe titive) at intermediate levels of physical stress where many basal species are able to tolerate environmental conditions and limit the availability of resources, and that positive (facilitative) interactions are more prevalent in either more physically har sh environments where neighborhood buffering maintains community structure, or less physically stressful areas where associational defenses play a significant role due to strong consumer pressure (Hay 1986, Bertness and Callaway 1994) . Accordingly, we predi ct that facilitation is the dominant interaction in multiple foundation species assemblages where the structure of a primary, stress tolerant foundation species creates a new, buffered habitat where other, obligate foundation species can proliferate and co mpetition among foundation species is most important where a number of species can act as primary space holders. Facilitation Where might facilitation be the dominant interaction among foundation species if they are, by definition, dominant, habitat formi ng organisms vying for space? The SGH predicts that facilitation should be prevalent in environments with strongly limiting physical factors, such as high evaporative or wave stress and low nutrient or water


27 availability (Bertness and Callaway 1994) , where the first foundation species to colonize a habitat experiences limited productivity and thus cannot completely dominate space. In stress maintained interstitial space, less tolerant foundation species may opportunistically proliferate due to the initial habitat modification by the first foundation species. For example, Acacia depanolobium trees can persist independently in dry Kenyan savannas although their productivity is restricted by a combination of stress factors, including low precipitation, fires and heavy browsing by grazers (Riginos 2009) . Once established, Acacias facilitate prairi e grass survival and productivity by locally reducing evapotranspiration, increasing water availability due to hydraulic lift, and enriching soil nutrients through litter fall (Belsky 1994) . Thus, by suppressing the dominance of the first foundation species to colonize a habitat and opening up space, elevated physical stress can organize communities into nested hierarchical assemblages where additional foundation species and their facilitative effects are obligately d ependent on the first foundation species to colonize a habitat. Furthermore, the spatial arrangement of foundation species in nested assemblages will likely attenuate through communities and drive predictable patterns in the distribution of associated orga nisms that tend to congregate where structural complexity and resource availability (i.e., crevice size, light, nutritional resources) are highest (Figure 2 1 ). The SGH also proposes that facilitation plays a critical role in structuring communities at t he opposite end of the environmental stress gradient where biological stress is high ( Bertness and Callaway 1994) . According to the Menge Sutherland community regulation model (Menge and Sutherland 1987) , the productivity and food web complexity of a community increase with decreasing physical stress, resulting in


28 elevated consumer pressure and stronger suppression of the abundance and distribution of primary prod ucers where physical stress is low. Facilitation emerges in foundation species that are functionally resistant to consumers because of their structural (e.g., calcium carbonate skeletons, thorns, fibrous/woody tissues) and/or chemical defenses (e.g., alkal oids, terpenoids, phenolics). Although facilitative interactions among refuge providing and refuge dependent species have been widely recognized [i.e., mangrove roots buffer predation on juvenile fishes (Beck et al. 2001) , unpalatable herbs protect palatable neighboring plants from livestock grazers (Anthelme and Michalet 2009) ], how associational defenses might influence coexisting foundation species has received little attention. In environments of low physical stress where consumer pressure a nd competition among dominant space holders are predicted to be intense (Menge and Sutherland 1976) , foundation species that are able colonize the habitat and exclude other space holders can support other foundation species that are more susceptible to consumption or inferior competitors by providing novel substrate for colonization. For example, in neotropical forests, tree limbs commonly provide substrate for structurally complex and productive epiphytic ferns and tank bromeliads that are poor competitors for space and vulnerable to graz ing and litter suppression on the ground (Matelson et al. 1993) but are key facilitators of arthropod communities in forest canopies (Ellwood and Foster 2004) . Similarly, in temperate macroalgae ecosystems, chemically defended, brown seaweeds, Sargassum filipendula and Padina vickersiae, provide essential substrate for the att achment of more palatable, epiphytic algae (i.e., Hypnea, Ulva, and Chondria spp. (Hay 1986) ) that secondarily host a range of invertebrates including sponges, tunicates, crabs and isopods. Within thes e structurally


29 dynamic tree epiphyte and seaweed algae communities, a rich fauna thrives (Hay 1986, Kitching 2001) despite high ambient consumer pressure in the ecosystem, illustrating how nested hierarchical foundation species assemblages may generate and maintain biodiversity hotsp ots within an ecosystem (Figure 2 1 ). Competition In hab itats experiencing intermediate levels of physical stress, competition, rather than facilitation, among foundation specie s tends to dominate because: the productivity of foundation species is higher, which limits the availability of unoccupied, interstitia l space for secondary organisms and therefore escalates competition among primar y space holders, and consumer pressure, still limited by physical stress, is too weak to mediate foundation species interactions. Where physical stress is strong enough to excl ude some potential foundation species but varies across the community, foundation species often are distributed in adjacent, monospecific zones where each is a primary space holder and maintains local dominance due to species specific trade offs in competi tive ability and stress tolerance. Such gradients in environmental stress, such as wave energy along coastlines or moisture and temperature variation across altitudes, are ubiquitous in natural environments and create an underlying basis for foundation spe cies segregation (Crain and Bertness 2006) . In New England salt marshes, for example, inverse gradients in inundation stress and nutrient availability segregate multiple foundation species ( Spartina alterniflora , Juncus gerardi , and Spartina patens ) in distinct zones parallel to shore that are determined by speci es specific trade offs in inundation tolerance and competitive ability (Levine et al. 1998) . Likewise, Patagonian rocky intertidal communities are organized by a competitive hierarchy where extensive bed s of the desiccation tolerant and competitively inferior


30 mussel, Perumytilus purpuratus, are displaced to physically stressful mid and high intertidal zones, while stress intolerant but competitively dominant coralline algae monopolize the low intertidal zone (Bertness et al. 2006) . These two foundation species differ in structural complexity (i.e., dimensions of interstitial space, thermal buffering) , which cascades up to other organisms, influencing the distribution and abundance of sea stars, limpets and crustaceans (Silliman et al. 2011) . In general, we predict that, as seen on Patagonian rocky shore s, spatial segregation of foundation species along stress gradients gives rise to variation in the composition of associated organisms that selectively congregate within particular foundation species a nd t heir functional traits (Figure 2 1 ). Thus, persiste nce of multiple foundation species that facilitate different suites of organisms is likely critical to maintaining overall species diversity and community stability. Effects of Foundation Species Diversity and Habitat Complexity on Biodiversity The int eractions among foundation species have cascading effects on the diversity and abundance of associated organisms. By providing a variety of refuge sizes, substrates and microclimates, multiple foundation species add multiple level s of habitat complexity (Altieri et al. 2007) which in turn mediates niche availability and predator prey and competitive dynamics. Whether foundati on species exist in nested or adjacent assemblages, however, will determine whether their combined affect enhances diversity locally by overlapping foundation species or cumulatively over larger scales when species distributions are segregated across space by adjacent foundati on spe cies, respectively (Figure 2 1 ). In addition, foundation species productivity will have important implications for the strength of habitat modification (species with larger


31 biomass should be stronger modifiers) and thus potential to facilitate other foundation species or associated organisms. The Ecological Theater and Evolutionary Play: the Context Dependence of Foundation Species Interactions We propose nested and adjacent assemblages as distinct types of hierarchical organizat ion that structure communities by facilitative or competitive interactions, respectively. In practice, however, interactions among foundation species are context dependent, varying spatially across landscapes (Wesenbeeck et al. 2007, Rigi nos 2009) and temporally over foundation species ontogenies or fluctuations in their productivity (McAuliffe 1984, Hay 1986) and, therefore, do not necessarily fall n eatly into these categories. In any ecosystem (e.g., savannas, cloud forests, coral reefs), a pool of potential space holders exists, and local biotic and physical conditions determine whether dominant species exclude the rest of the pool or whether stress tolerant species take hold and facilitate the growth of others. Consequently, the same foundation species that interact to form nested assemblages under some conditions may be organized through competitive hierarchies under different conditions. Case exam ple 1 Hydrodynamic forces (wave exposure and wind stress) vary along the New England coastlines, and it has been shown in field experiments that they drive predictable patterns in the strength and direction of the interactions among primary space holders (Wesenbee ck et al. 2007) . Under high hydrodynamic stress, vegetated cobble beach communities prevail in which stress tolerant cordgrass positively interacts with secondary space holders (e.g., forbs, grasses, sedges) to sustain diverse plant communities. Along wav e protected coasts, however, competitive interactions


32 predominate, and inferior sedges and forbs become excluded by stress intolerant but competitively superior grasses. As a result, distinct community types with unique spatial structures (e.g., vegetated cobble beach, fringing marsh, salt marsh) arise across hydrodynamic gradients because of variation in the strength and nature of the interactions among foundation species, despite a common spe cies pool across all habitats (F igure 2 2 ). Case example 2 In t he heavily grazed, arid Nigerian Sahara, interactions between Acacia tortilis var. raddiana , a leguminous tree that improves soil nitrogen availability and provides protection against soil erosion, and the dense nurse tussock Panicum turgidum , which facili tates a diverse assemblage of desert forbs and herbs (Anthelme and Michalet 2009) , shift from a facilitative to a competitive interaction as Acacia trees progress from early life stages to mature adult trees. As seedlings, Acacia growing alone are browsed intensively by livestock, but those that germina te within the complex matrix of grazer resistant Panicum often persist and continue to grow, which suggests that the associational defenses provided by the tussocks are critical to the establishment of this economically important foundation species (Anthelme and Michalet 2009) . As Acacia mature and beco me less vulnerable to grazers, they compete with Panicum for water, however, and can limit grass biomass (Ludwig et al. 2004) . In this example and potentially many other ecosystems in which consumers selectively browse young and vulnerable species, the associational defenses provided by grazer resistant foundation species may be essential to the long term stability of the community and a provision of key ecosystem services.


33 Since the physical and biotic environment can moderate the strength and even reverse the direction of foundation species interactions, identifying the environmental context in which foundation species coexist is of central importance to predicting the nature of the communit emphasizes that climate change, which may alter environmental gradients or the anthropogenic modification of natural grazer regimes through activities such as overfishing or intensive livestock g razing, can fundamentally alter the dynamics among habitat forming dominant species, with cascading effects on dependent organisms. Conclusions Decades of research have emphasized biogenic habitat creatio n by single foundation species (Dayton 1972) . Here we revise this approach to understanding community organization by developing the idea that multip le foundation species interact in most ecosystems to synergistically structure communities, enhance species diversity and stabilize ecosystem function. We suggest that future investigations of community assembly need to consider the hierarchical organizati on of foundation species, the strength and direction of their interactions and the structural complexity of habitats that arise from their presence, persistence and interactions. Additionally, acknowledging that adjacent assemblages, that historically have been studied in isolation, may be a predictable outcome of interactions among foundation species exposes the need to adopt landscape scale perspectives of communities and underlying stress gradients. Loss and degradation of foundation species due to defor estation, pathogens, depletion of top predators, urban development, climate change, and eutrophication is a widely recognized global problem (Coleman a nd Williams 2002, Ellison et al. 2005, Altieri and Witman 2006, Bracken et al. 2007) . In addition to strategies to prevent further


34 loss of foundation species, we suggest a more proactive approach to restoration that prioritizes re establishing foundation species assemblages to degraded ecosystems as a means to restore stable, diverse communities. In Nigerian Sahara, for example, intensive livestock browsing is linked to low regeneration of the key leguminous tree, Acacia tortilis (Anthelme and Michalet 2009) . A recent study has demonstrated that transpl anting Acacia seedlings within the protective matrix of the naturally abundant nurse tussock, Panicum turgidum , increases tree survivorship and growth at vulnerable, early life history stages. As a result, Anthelme and Michalet (2009) recommend that future conservation efforts utilize natural grazing refuges (i.e., the foundation species, Panicum ) as a cost effective solution for enhancing the density of Acacia in an effort to restore biogenic structure and ecosystem functioning in these landscapes. Likewis e, the success of seagrass restoration projects in fostering the return of plants and fauna to soft sediment habitats in Chesapeake Bay, USA might be improved if multiple seagrass species, rather than only Zostera marina , are seeded because it would result in the re establishment of complex community structure over r elatively short time scales (Marion and Orth 2008) . In general, harnessing facilitation (Byers et al. 2006, Halpern et al. 2007) among naturally synergistic foundation species, not just the presumed competitive dominant, by planting or seeding should be evaluated as a tool to restore biological communities and ecosystem services in severel y degraded habitats.


35 Figure 2 1. Types of foundation species assemblages . Foundation species may form A ) nested or B ) adjacent assem blages. In nested assemblages , positive interactions hierarchically structure communities in facilitation cascades where the first foundation species to colonize a habitat facilitates other foundation species and through complementary structural complexity they support diverse species assemblages. In adjacent assembl ages , foundation species compete for space to form discret e competitively determined zones, where structural complexity is mediated locally by the dominant foundation species and drives variation in community composition across zones. In both nested and adjacent assemblages, multiple foundations species are neede d to support diverse communities and maintain higher order interactions at landscape scales. Nested assemblages are apparent in the Sonoran Desert where shading and nutrient deposition by nurse shrubs fac ilitate the growth of saguaro C) , and the adjacent a ssemblages of salt marsh grasses are a striking feature along wave pro tected New England shorelines D ). Photo credits: Joe Shaw and Andrew Altieri.


36 Figure 2 2. Context dependence of interac tions among foundation species. A) Despite a common pool of spe cies, sho reline community composition varies widely across hydrodynamic stress gradients in New England as a consequence of foundation species interactions shifting from competitive along wave protected coasts, to facilitative along wave battered cobble be aches. B) Likewise, the nurse tussock, Panicum provides a critical refuge for Acacia seedlings from grazers in intensively browsed, arid Nigerian landscapes, but competes with Acacia for water as trees mature and become less vulnerable to predation . Photo credits: Andrew Altieri and Fabien Anthelme.


37 CHAPTER 3 SECONDARY FOUNDATION SPECIES AS DRIVERS OF TROPHIC AND FUNCTIONAL DIVERSITY: EVIDENCE FROM A TREE EPIPHYTE SYSTEM Introduction Experimental and comparative studies of habitat forming foundation spe cies (sensu Dayton 1972, see Bruno and Bertness 2001 for refined definition) have revealed time and again that these organisms can have positive effect s on biodiversity by generating the structure and conditions within which other species and their interactions, such as competition or predation, occur (Bruno & Bertness 2001; Stachowicz 2001; Bruno et al. 2003) . The overwhelming focus of this research has bee n on single foundation species [ e.g., Douglas fir ( Pseudotsuga menziesii ) , Poplar ( Populus angustifolia and P. fremontii ), smooth cordgrass ( Spartina alterniflora ) ] or functional groups (e.g., nurse shrubs, corals, kelp) because these spatially dominant organisms are thought to be the primary drivers of community composition (Whitham et al. 2006, Irving and Bertness 2009) . Secondary, or dep endent, foundation species can be among those organisms facilitated b y a primary foundation species [ e.g., ribbed mussels ( Guekensia demissa ) within cordgrass] and, by creating significantly more structure or unique refuges from physical or biotic stress, may modify the number and identity of individuals in the local community (Altieri et al. 200 7, Yakovis et al. 2008b, Bishop et al. 2012, Dijkstra et al. 2012) . Given that multiple habitat forming species are present in many systems, secondary foundation species may be common, but currently underappreciated, drivers of biodiversity and ecosystem functioning (Angelini et al. 2011, Thomsen et al. 2013) . Ecologists interested in whole community facilitation by foundation species (Bruno and Bertness 2001, Whitham et al. 2006, Rowntree et al. 2011) have


38 documented higher abundance and richness of associated species where primary foundation specie s facilitate secondary foundation species (Altieri e t al. 2007, Yakovis et al. 2008, Bishop et al. 2012, Dijkstra et al. 2012) . This positive direct interaction among foundation species that gives rise to indirect facilitation of biodiversity has been coined a facilitation cascade (sensu Altieri et al. 2007) . To date, no study has ex perimentally tested for the presence of fa cilitation cascades in terrestrial or aquatic systems impacts beyond abundance and richness measures to include more functionally informative metrics of community structure (see Altieri et al. 2007, Dijkstra et al. 2012) . To gauge whether the facilitation cascade concept holds broad utility, more re search on Likewise, analyses that distinguish whether secondary foundation species simply enhance the size (i.e. support more individuals of the same species, feeding guilds or life stages) , and/o r influence trophic structure (i.e. support new species and feedi ng guilds) and life stage diversity (i.e. support juveniles) of local communities are critical to evaluate whether facilitation cascades also drive spatial patterns in food web complexi ty, sp ecies interactions, and nursery habitat community characteristics that influence ecosystem functioning ( e.g. Hooper et al. 2005) . Beyond understanding the generality of facilitation cascades and their impact on community structure, we must also begin to construct conceptual models that formalize predictions for when and where these hierarchical chains of positive interactions will generate hot spots of biodiversity and the more numerous and complex species interactions and ecosystem process es that can follow.


39 The association between trees and vascular epiphytes is a useful system in which to explore generality of facilitation cascades and the contribution of primary and secondary foundation species in regulating trophic structure and ecosyst em function . Epiphytes, including thousands of charismatic species from Bromeliaceae, Orchidaceae, and Araceae families , are distributed throughout tropical and sub tropical latitudes (Benzing 1990) and generate intricate structures within the broader architecture of host trees (Fr eiberg 2001, Stuntz et a l. 2002 ) . Although it is commonly assumed that epiphytes are intolerant of abiotic (e.g., high moisture, low light) and biotic (e.g. high consumer pressure) conditions on the ground (Nadkarni 1992, Matelson et al. 1993, Mondragn et al. 2004) , ecologists have yet to use mani pulative experiments to test if and how epiphytes depend on host trees (Zotz and Hietz 2001) . In addition , while epiphytes are commonly recognized as facilitators in metacommunity, tropical biodiversity, and community genetic studies [e.g., they often form complex structure s that support other organisms such as ants, frogs, protists, snails, spiders, and birds (Nadkarni 1989, 1994, Ellwood and Foster 2004, Dial et al. 2006, Cruz Angón et al. 2009, Zytynska et al. 2011, Yanoviak et al. 2011) ], they are not typically identified as foundation species (but see Thomsen et al. 2010 for discussion) and never been studied as critical, intermediate links in facili tation cascades. Here I explore the mechanisms that generate and community level consequences of a facilitation cascade that likely organizes the most conspicuous tree epiphyte assemblage in the southeastern US: Southern live oaks ( Quercus virginiana , he reafter, oaks) laden with the atmospheric bromeliad, Tillandsia usneoides (commonly Spanish moss, hereafter Tillandsia , Figure 3 1 ). Specifically, I test: whether oaks and


40 Tillandsia can act as primary and secondary foundation species, respectively, and co llectively generat e a facilitation cascade, and if Tillandsia enhances the density and diversity of species, feeding guilds, and life stages and thus supports communities that are larger than and functionally distinct from those associated with oaks alone. I then compare invertebrate communities associated with oaks and both oaks and Tillandsia at sites distributed across the southeastern US to test whether this epiphyte supports more complex food webs and generates nursery habitat wherever it overlaps with oaks. Methods Field experiments were conducted in the National Estuarine Research Reserve on Sapelo Island, GA, USA (31°24'2"N, 81°17'4"W) in Bahia grass ( Paspalum notatum ) savannas intersperse d with open grown (>30m diam. ) oaks. This oak species was se (Callaway et al. 2002) . Tillandsia , a CAM photosynthesizing, rootless vascular plant that forms high abundance and widespread distribution in the southeastern US (Benzing 1990) . In this region, many other tree species host Tillandsia (Callaway et al. 2002) and other epiphyte species colonize oaks (Benzing 1990) . Oak F acilitation of Tillandsia : an E xperiment We first assessed whether Tillandsia depends on primary facilitation of trees by tracking the survival of festoon s left on and removed from oak hosts. In May 2010, we identified a large oak at two sites and, within each oak, selected 15 festoons, standardized for volume (40 cm 3 ), health (90 95% live tissue, see below), and position (4m off ground, 5m from canopy edge ) on different branches and randomly assigned each to one of three treatments: oak limb control, procedural control, or oak limb


41 removal [N= 5 replicates per treatment per site]. For oak limb control treatments, we flagged but did not alter the position of festoons. For procedural control treatments, we removed and immediately replaced festoons on oak limbs to account for disturbance effects. For oak limb removal treatments, we removed and dropped flagged festoons on the ground. After 60 days, we assessed s urvival by saturating festoons in a bucket of water, refreshed after each replicate, and scoring plants for % live tissue. Wetting distinguishes live Tillandsia tissue, which is green when wet, from dead tissue , which is brown (Figure 3 2 ). P ercent surviva l data were arcsine (square root (X)) transformed to meet the assumptions of normality. We treated Site as a random factor, Oak Treatment as a fixed factor, and assessed their effect on festoon survival using mixed effects Analysis of Variance (ANOVA). We then tested the hypothesis that shading and elevation off the ground are two mechanisms by which host trees facilitate Tillandsia. To do so, we haphazardly assigned a shade and elevation treatment to each of 40, 40 cm 2 plots (4 treatments: shade/no shade elevated/ ground, N= 10 replicates ) positioned within a Bahia grass patch located 75m from a Tillandsia laden oak in July 2010. We then air dried and weighed 30g Tillandsia transplants, collected from a single oak and standardized for live tissue (90 95% ). We secured 2 layers of black mesh to corner stakes 60cm above shade plots and left no shade plots unmanipulated. To assess if shades mimicked the ~60% light attenuation provided by oak foliage in savannas (Figure 3 3 ), we monitored Photosynthetically Ac tive Radiation (PAR) in shade and no shade plots on a cloudless day using a handheld light meter. To manipulate elevation, we positioned transplants 25cm off the ground on grass turfs in elevated plots or directly on the ground in


42 unmanipulated ground plot s. We placed an iButton (Embedded Data Systems , Lawrenceburg, KY), programmed to record one temperature reading per hour, in one we monitored transplant survival as abo ve. The effect size and significance of Shade, Elevation, and their interaction over time on the percent Tillandsia survival , transformed as above, was assessed with repeated measures ANOVA. Tillandsia Facilitation of Invertebrates: the Physical E nvironmen t To assess whether Tillandsia functions as a secondary foundation species that moderates physical stress within oaks, we monitored temperature and humidity, factors known to influence invertebrate desiccation and survival (Wigglesworth 194 5, Stuntz et al. 2002, Ellwood et al. 2011) . In one oak at two savannas, we deployed 3 pairs of Hygrochron iButtons (see above), programmed to record temperature and relative humidity hourly from April 7 May 17, 2012. In each pair, one logger was positio ned on a Tillandsia colonized limb surface and the other 0.5m away on a limb of a similar size and elevation that lacked Tillandsia . This arrangement was used to minimize variation in canopy microclimat e and thus isolate the e ffect of Tillandsia. For each pair, we calculated the difference (i.e., Y i, = Tillandsia colonized i, un colonized i ) in mean daytime temperature, mean humidity, and the coefficient of variation (CV) of humidity. We assessed the CV in humidity because invertebrates have been shown to su rvive better in habitat s characterized by stable ( low CV) humidity conditions (Bertrand & Wilson 1996) . We then calculated the mean difference for each metric for each day at each site [Y m = (Y 1 + Y 2 +Y 3 )/ 3] and used a t te st to assess if mean differences in each metric significantly differed from zero .


43 Tillandsia Facilitation of Invertebrates: S urvival To assess if Tillandsia is an effective foundation species and improves invertebrate survival within oaks, we conducted tw o experiments. First, to assess whether Tillandsia increases invertebrate survival in the absence of predators and alleviates other potential sources of mortality within oaks, we constructed 30 × 60 cm mesocosms (Diam. × L, Figure 3 4 ) and stocked each wit h either a 25 cm limb draped with a vacuumed 25 g Tillandsia festoon or only a 25 cm limb (N=3 replicates per treatment ). We attached a n iButton, programmed as above, to each experimental limb and added 10 juvenile isopods ( Venezillo parvus ) that are commo n within oaks at our study site to each mesocosm. Mesocoms were hung 4m off the ground from limbs with similar light exposure within one oak. After 7 days, we collected the iButtons, counted live isopods, and used a generalized linear model with a quasibin omial error function to test the e ffect size and significance of Tillandsia Treatment on th e number of live and dead isopods in R version 2.15.1 (R Core Development Team 2012) . Next, to test if Tillandsia modifies predator foraging efficiency to enhance prey survival, we added either a 25 cm long oak limb and 25 g festoon , vacuumed to remove invertebrates, or only a 25 cm l imb to 5L arenas (N=5 replicates per Tillandsia treatment ) positioned in the shade and stocked each with 4 juvenile field crickets ( Gryllus spp.) . We allowed crickets to acclimatize for 2 hours before adding one spider ( Gladicosa pulchra ) to each arena. C r ickets and spiders were used as representative prey and predators because they a re common and abundant in our study system . After 12 hours, we counted live crickets and inspected dead crickets for spider wounds to rule out ot her potential causes of death. The effect of Tillandsia Treatment on the number of live and dead crickets was analyzed with a generalized linear model as above.


44 Oak Tillandsia Facilitation of I nvertebrat es: Community R esponses To test the hypothesis that Tillandsia not only increases invertebrate abundance and richness but also enhances food web structure (i.e. increases the number and diversity of feeding guilds) and functions as a nursery (i.e. supports juveniles) within oaks, we identified 2, 1 m 3 festoon colonized plots standardiz ed for limb diameter, height off the ground (4m), festoon volume (30 30 60cm, L W H), festoon health (90% live), and distance to adjacent festoons (2 3m to nearest neighbor), in each of 15 oaks. In each oak, we randomly assigned a Tillandsia remova l or control treatment to each plot: in removal plots, we extracted the festoon, leaving only the oak limb and, in control plots, we removed and immediately replaced the festoon. After 8 weeks, we used a 50 gallon plastic bag to envelop each plot, collect the festoon (if present), and capture all detritus and invertebrates that we brushed from each limb surface for 30 seconds. We ran each bagged sample over a vacuum (i.e., leaf blower equipped with an insect screen attachment) for 1 minute to isolate invert ebrates. All macro invertebrates were counted and sorted to morphospecies (hereafter, species) , life stage (juvenile or adult, b ased on size and genitalia development ), and feeding guild (predator, parasite, scavenger, detritivore, folivore, granivore, or nectavore assigned after consultation with experts at Florida Department of Agriculture and Consumer Services, Division of Plant Industries). T o evaluate Tillandsia on different metrics of community structure, w e assigned Tree as a random factor, Tillandsia Presence as a fixed factor, and used a mixed effects ANOVA to assess their effect size and significant on invertebrate density (total # ind. per m 3 ), juvenile density (# juv. per m 3 ), species richness, sp ecies diversity


45 (Shannon Wiener, . To investigate whether the composition of functional groups differed in Tillandsia present and removal plots , we calculated the proportional contribution of each functional group and conducted a mul tivariate ANOVA using the lme4 package in R (Bates et al. 2012) . Significance of Tillandsia (as in Jaschinski et al. 2009) . We also rarefied our data to compare species and feeding guild richness metrics at a standard sample size (Gotelli and Colwell 2001) using the rich pa ckage in R (Ro ssi 2011) . To evaluate the significance of Tillandsia treatment on rarefied richness metrics, we conducted 4 99 randomizations of our data, drawing samples with replacement, to calculate the probability, p, of observing difference s between Tillandsia pres ent and removal plots e qual to or more extreme than observed difference s in richness (Rossi 2011) . G enerality of O ak Tillandsia Facilitation of Invertebrates: Latitudinal S urvey In September 2010, we surveyed invertebrate communities associated with oak or both oak and Tillandsia in the strength of this potential facilitation cascade across a range in which live oaks and Tillandsia commonly overlap. At each site, we identified 6 oaks [only 5 in Palm Coast (FL1) because of time constraints] colonized by Tillandsia and, within each, selected one 1 m 3 un colonized ( Tillandsia absent) limb plot and one 1 m 3 Tillandsia coloniz ed limb plot, and bagged, brushed, and vacuum sampled invertebrates. Sampled invertebrates were stored in 70% ethanol. Using a microscope at 25 × , we counted macro and micro invertebrates and categorized individuals as above. Since we had no a priori expec tation that communities would differ among trees within a site, we


46 excluded tree as a factor , nested Tillandsia Presence with in S ite , and assess ed the effects of Site and Presence [Site] on the community metrics listed above using ANOVA. We used multivaria te ANOVA to test the significance of Site and Presence Within each site, we used randomization tests to calculate the probability, p, of measuring species and feeding gui ld richness differences between Tillandsia present and absent plots equal to or more extreme than our observed difference s . Results After 2 months, festoons assigned to oak control and procedural control treatments positioned on limbs su rvived ( >85% liv e tissue ) , while those removed from trees (i.e. , on the ground) died 2, 26 = 357.6, P<0.0001, Figure 3 5 A ). In the experiment designed to test mechanisms of tree facilitation , Tillandsia positioned in control light and no elevation ground plo ts experienced intensive light and p ers istent ly high temperatures (Figure 3 5 B ) and died rapidly (Fig ure 3 5 C ). In contrast, Tillandsia survived for the 56 day experiment in treatments where transplants were both shaded and elevated (Ti me * Shade * Elevation: F 5,32 = 20.7, P = 0.0007, Figure 3 5 C ). Monitoring revealed that daytime temperatures were typically cooler on Tillandsia colonized relative to paired un colonized limbs at our two sites ( Tillandsia P resence: t 40 > 8.6, P<0.0001, Figure 3 6 A ]. In addition, relative humidity was higher ( Tillandsia P resence: t 40 > 12.6, P<0.0001 Figure 3 6 B ) and more stable (i.e., lower CV, Tillandsia Presence: t 40 > 4.18, P < 0.0001, Figure 3 6 C ) on Tillandsia colonized relative to un colonized limbs.


47 In experimental mesocosms hung within an oak, Tillandsi a incr eased juvenile isopod survival (77 ± 9 vs. 37 ± 9% alive, Tillandsia presence: T 4 = 2.81 , P=0.0 482). Temperature was lower and humidity higher in Tillandsia present rel ative to absent mesocosms (Figure 3 4B,C ) , as obse rved in canopy monitoring (Figur e 3 6 ). Similarly, Tillandsia significantly reduced spider predation to increase juvenile cricket survival [95 ± 5 (mean ± SE) vs. 60 ± 6 % alive, Tillandsia presence: T 8 = 2.74, P= 0.0256 ) . All dead crickets had spider wounds. In our field experiment, all community metrics were higher in Tillandsia present than removal plots ( Table 3 1). Tillandsia present plots not only contained 20 more invertebrates, but also functioned as a nursery and supported 60 more juveniles than Tillandsia removal limbs ( juveni les comprised 30 v s. 2% percent of the total community, on average , in each plot type ). Similarly, Tillandsia increased species richness and and 10 fold and 11 fold relative to removal plots. The compositi on of functional groups was also significantly different among treatments, such that d etritivores, omnivores, predators, and parasites were common in Tillandsia p resent plots, while only detriti vores were common in Tillandsia re =0.61, F = 7.50, P= 0.0002). Rarefaction indicated there were ~ 2 more species and feeding guilds on Tillandsia present than removal limbs. In our survey, we found a significant effect of site on invertebrate density, richness, and diversity ; the magnit ude of between site differences was minor relative to that of Tillandsia presence effects within site, however ( Figure 3 7 , Table 3 2, 3 3 ). > 16, 5 and 1.7 × higher in Tillandsi a pre sent plots than absent plots . We found no juveniles in un -


48 co lonized limbs at 4 sites and , at the 2 sites where they were found in both plot types, Tillandsia increased average juvenile density from less than 1 to more than 80 individuals per m 3 . Feeding g uild richness and were > 2.4 and 1.5 × greater on Tillandsia present relative to absent limbs. Invertebrates were largely detritivores and scavengers on bare limbs , while those on Tillandsia present limbs often included det ritivore (e.g., roaches, isopo ds , scale bugs), scavenger (e.g., ants , crickets ), granivore (e.g., weevils), predator (e.g., ladybird beetles , spiders ), and parasite (e.g., mite, parasitic wasp) guilds ( Table 3 2, 3 3 ). Finally, rarefaction indic ated communities were more species and feeding guild rich on Tillandsia present versus absent limbs at 5 of our 6 sites (Figure 3 7 ). Discussion Our results demonstrate a facilitation cascade structures oak Tillandsia invertebrate communities in southeas tern US coastal savannas. In this assemblage, oaks function as primary foundation species that ameliorate solar and temperature stress with their foliage and elevated limbs to facilitate Tillandsia . Tillandsia, in turn, functions as a secondary foundation species within oaks and further reduces desiccation stress and predator foraging efficiency to facilitate invertebrate s. Importantly, our analyses reveal that Tillandsia supports communities that are not only larger and more species rich but also contain f ar more juveniles and feeding guilds than those supported by oaks alone . Together with examples from marine and coastal systems, this study advances two ideas : hierarchical organization of communities based on direct and indirect facilitation of multiple f oundation species occurs across biomes, and secondary foundation species can generate addition habitat and unique refuges


49 within primary foundation species to locally enhance species abundance, food web complexity, and key ecosystem functions, such as nurs ery provisi on. Mechanisms Maintaining Tree Epiphyte Facilitation C ascades Although it is intuitive that host trees facilitate epiphytes (Zotz and Hietz 2001) , the mechanisms underpinning this obligate assoc iation have never been tested experimentally as far as we are aware. Studies that monitor the fate of dislodged epiphytes infer from observations that physical (e.g., low light, high moisture) and biotic (e.g., consumption by invertebrates, bacteria, and f ungi) pressures not present in tree canopies are the likely drivers of plant death on the ground (Nadkarni 1992, Matelson et al. 1993, Zotz and Hietz 2001) . In our experiment s , Tillandsia senesced when removed from host trees (Figure 3 5 A ) suggesting it is intolerant of temperature and light levels associated with the ground in the temporal and spatial context of our study, but survive d for 2 months in the terrestrial environment if the se specific physical stressors we re relieved (Figure 3 5 C ). In other seas ons, habitat types , or latitudes, we suspect that Tillandsia depends on host trees for refuge from a number of other stress ors , such as high moisture, given that its trichomes flatten and block CO 2 exchange when wet (Martin and Siedow 1981) , nutrient lim itation, since it acquires nutrients from tree leachates and atmospheric sources (Benzing 1990, Zotz and Hietz 2001) , freezing temperatures, and fungal infestation (CA, personal observation ). The broad conclusion here, that primary foundation species reduce multiple stressors to facilitate secondary foundation species, is consistent with other facilitation cascade studies (Altieri et al. 2007, Bishop et al. 2012) and highlights the fundamental importance of primary foundation species: without their initial generation of habitat and modification of environmental conditions, most


50 other species are unable to establish and interactions among those species, including additional facilitation, fail to arise (Bruno and Bertness 2001, Silliman et al. 2011) . Within oaks, we found that festoons not only increase the amount of structure (Figure 3 1) , but also moderate two physical stressors , temperature and humidity (Figure 3 6 ). Relative to bare limbs, Tillandsi a can also extend periods of significantly lower temperatures and higher humidity for 4 5 days longer after rainstorms, which occurred during a trial iButton deployment (CA, unpublished data ), but not during the persistently dry 2012 monitoring period. Pri or studies have shown that elevating temperatures only a few ºC can increase water loss from invertebrate larvae and pupae by an order of magnitude (Wigglesworth 1945) ; by inference, we s uspect that the more favorable physical conditions with in Tillandsia may be key to invertebrate survival. In fact, our results indicate isopods suffer higher mortality on drier, hotter un colonized limbs compared to moister, cooler Tillandsia colonized li mbs. Although this simple experiment did not test specific mechanisms by which Tillandsia facilitates invertebrates, it did show that Tillandsia and oak limbs generate a higher quality habitat for isopods than oaks alone . Since many epiphytes, such as tank Nest ferns, form large structures that modify moisture and temperature levels within trees (Nadkarni 1994, Ellwood and Foster 2004, Dial et al. 2006) , we predict they also fun ction as secondary foundation species and support novel species, life stages, and feeding guilds that require buffered environmental conditions. Our mesocosm experiment ind icated Tillandsia also affects the strength of predator prey interactions and thus a meliorate s biotic stress. Given the relative size of spiders (larger) and juvenile crickets (smaller), we suspect that crickets maneuver


5 1 through the intricate structure of festoons more easily than spiders and therefore better escaped predation in Tillands ia addition than absent mesocosms (Persson 1995, Dijkstra et al. 2012) . The outcome of this experiment could potentially be reversed if other predator or prey species were considered: Synema viridans , the green crab spider, for instance, is a cryptic ambush predator that is common within Tillandsia (CA, personal observation ) and may catch prey more efficiently in festoons relative to bare limbs. The concentration of p otential p rey within Tillandsia also seemed to attract pred ators (e.g., up to 7 and 15 spider species and individuals, respectively, in a single festoon), which may further influence the frequency and outcome of predator prey encounters (Crowder and Cooper 1982) . Regardless of whether Tillandsia increase s the relative success of a given predator or prey, our result s support classic studies that identify habitat structure as a key factor controlling of species interaction strength (e.g., Crowder and Cooper 1982, Irlandi 1994, Persson 1995) . Given that habitat structure is generated by foundation species that themselves evolve, grow, and interact with the environment and other species , as opposed to static non living or generically labeled the relationship between habi tat structure and species interaction strength is certainly more dynamic than often considered (Bruno and Bertness 2001) . In our study system, for example, the architecture of o ak limbs and morphology of Tillandsia festoons vary across savanna and forest ecotones and with plant age, features of this living habitat that likely elicit changes in both community composition and st rength of species interactions.


52 Secondary Foundation S pecies as D rivers of Trophic Structure and Ecosystem F unctioning Results from our removal experiment and latitudinal survey suggest that Tillandsia does more than simply increase the number of invertebrate individuals: instead, it facilitates a greater num ber of species , feeding guilds , and life stages across its range of overlap with oaks (Table 3 1, Table 3 2, Figure 3 7 ) . Although patterns in community structure were similar between our experiment and all surveyed sites, inclusion of super abundant, tiny organisms in our survey counts modified our initial perception of Tillandsia the experiment in two ways: since micro invertebrates form a large portion of Tillandsia associated communities, we likely underestimated Tillandsia facilita tive effects by only includ ing macroscopic species, and since invertebrate density was ~100 , rather than 20 , greater on Tillandsia limb plots, we likely overestimated differences in rarefied richness between plot types. In other words, our survey reveal ed that Tillandsia communities are larger, more species rich, and less even than measured in the experiment. In addition, Green Anole lizards ( Anolis carolinensis , CA, personal observation ) , Cuban tree frogs ( Osteopilus septentrionalis , PDR. Silliman, pers onal observation ), and Northern Yellow bats ( Lasiurus intermedius , C. Bland, unpublished data ) are often found in festoons , hinting the impact of this facilitation cascade extends to larger organisms. Interestingly, folivores, which derive energy and nutri ents from live plants, were rare in all of our samples, while detritivores, which assimilate resources from decomposing plant s and transfer them to higher trophic levels, w ere common and abundant . D ominance of detrit ivores has been noted in many epiphyte s tudies (Nadkarni 1994, Ellwood and Foster 2004, Dial et al. 2006, Cruz Angón et al. 2009) and may suggest that brown


53 ( detritus b ased), rather than green ( plan t based), food webs mediate energy flow in systems with high epiphyte cover . Tillandsia also appears to function as a nursery. In the field, we often noted eggs, molts, and nests in festoons and, in our survey samples, 3,512 of the 3,515 juvenile inverteb rates including spiderlings, cricket and cockroach nymphs, and larval ants, counted were from Tillandsia present plots. Ot her f acilitation cascade studies have documented higher densities of juveniles within secondary foundation species as well (Altieri et al. 2007, Bishop et al. 2012, Dijkstra et al. 2012) . Intuitively, the concentration of early life stages within secondary species makes sense: thes e organisms are often susceptible to physical and biotic stress and t herefore benefit from the large and small structural traits of primary and secondary foundation species that provide multipl e levels of stress amelioration . For the past decade , whole community facilitation by foundation species has been recognized (Bruno and Bertness 2001, Bruno et al. 2003, Whitham et al. 2006, Rowntree et al. 2011) , and w hat our study and others on facilit ation cascades contribute to this concept is the idea that secondary foundation species may complement and magnify the e ffects of primary species on trophic structure and key ecosystem functions, such as pollination, nutrient cycling, and pest control, by support ing particularly vulnerable species, life stages, a nd feeding guilds. Foundation Species Biodiversity (FSB) Model To date, facilitation cascade studies have focused on examples where layering of secondary foundation species within primary foundation species powerfully enhances species abundance and divers ity . In many systems where they occur, facilitation cascades may have less dramatic effects on some, or all, metrics of community structure, however, because the secondary foundation species negligibly, or to a lesser


54 extent, influences the amount of habit at that is available (e.g., they are uncommon or small) or generates habitat that is functionally redundant with that of the primary (e.g., they exhibit similar structural attributes). In the Foundation Species Biodiversity (FSB) model (Figure 3 8 ), we in tegrate this context dependence and hypothesize that the size and functional traits of secondary foundation species, as the key, intermediate links in facilitation cascades, control their relative impact on associated sp ecies abundance and diversity. Speci fically, the FSB predicts that where secondary foundation species increase habitat availability, they provide the substrate and resources necessary to support the settlement, growth, and /or retent ion of more individuals of species that already occur in the ) . A Type A facilitation cascade might occur , for example, where manatee grass ( Syringodium filiforme ) secondarily establishes within a meadow of structurally similar shoal grass ( Halodule wrightii ) in shallow marine environments and, i n increasing shoot density, significantly enhances the abundance, but not richness, of resident invertebrates and fish. Likewise, the FSB predicts that where secondary foundation species exhibit functional traits (e.g., surface texture, crevice size, chemi cal composition) that are very different from those of the primary foundation species, their generation of novel refuges from physical (e.g., via shading, baffling wave or wind stress , stabilizing humidity , providing hard substrate ) or biotic (e.g., via fo rming small er or differently shaped crevice s for prey) stressors facilitates the settlement and retention of new species, life stages, and feeding guilds An example of a Type B facilitation cascade might occur where pen shells ( Atrina rigida ) s ettle sparsely within turtle grass ( Thalassia testidinum ) monocultures and, in providing stable


55 substrate for sessile invertebrates and complex shelter for egg laying fish, enhance the species, life stage, and functional diversity of the resident community but do not significantly alter the number of individuals. In following, where secondary foundation species increase habitat availability and exhibit functional traits different from those of the primary foundation species , like Tillandsia within trees (th is study) , mussel bed s within cordgrass (Altieri et al. 2007) , or algae and oysters layered on mangroves (Bishop et al. 2012) , they are predicted to increase both abundance and div ersity . Alternatively, where secondary foundation species do not considerably increase habitat availability or functional tr ait diversity (bottom left, Figure 3 8A ), like encrusting lichens on trees, the FSB predicts that although they will incr ease species abundance and diversity some, their facilitative effects will be small relative to those of the primary foundation species. Finally, this model may also be used to predict changes in community size and diversity over time and across environmen tal stress gradients as secondary foundation species grow or modify their functional traits in response to local conditions. In the future, experiments that manipulate the size and trait diversity of foundation species and use informative metrics to assess community responses are necessary to test whether the FSB predictions about where foundation species have maximal effects on biodiversity are indeed correct. Along with empirical studies, efforts to incorporate the powerful effects of foundation species i nto food web and meta community models will be essential in determining the mechanisms that cause community structure within foundation species to change over space and time. In closing, advancing our understanding of interactions among foundation species and


56 context dependency in facilitation cascades is an exciting frontier for ecological research and theory, with important implications for conservation .


57 Table 3 1. Invertebrate response to Tillandsia . Summary of invertebrate community response to Tillandsia presence or removal. Data are shown as the mean (standard error) for 15 replicates per treatm ent. 0.01 0.01 * Mean, standard error, and p value are based on 99 randomizations of Tillandsia present and removed plot communities conducted in R .


58 Table 3 2. Effects of Tillandsia acr oss latitudes. Results from nested Analyses of Variance testing for the effect of Site and Tillandsia Presence within Site on six metrics of invertebrate community structure. Response Site Tillandsia Presence [Site] (per m 3 plot) F 5,59 P value F 6,59 P value # Individuals 3.7 0.056 33.57 <0.0001 # Juveniles 0.84 0.53 66.33 <0.0001 Species richness 3.97 0.0036 47.54 <0.0001 diversity 0.89 0.49 6.13 <0.0001 Feed. guild richness 2.99 0.02 22.03 <0.0001 0.77 0.57 3.43 0.005 7 Table 3 3 . R esults from latitudinal survey rarefaction analyses. R esults from randomization tests performed for each site assessing the effect of Tillandsia Presence on rarefied species and feeding guild richness. Response variable Tillandsia Presence (P value from r andomization test) FL1 FL2 GA SC1 SC2 NC Rarefied Species Rich . . 0.01 0.02 0.01 0.01 0.01 0.01 Rarefied Feed. Guild. Rich . 0.07 0.01 0.04 0.01 0.01 0.01


59 Figure 3 1 . Live oak draped with Spanish moss. A potential example of a fac ilitation cascade is Quercus virginiana , Southern live oak, laden with Tillandsia usneoides , Spanish moss, in southeastern US coastal savannas.


60 Figure 3 2 . Method for scoring Tillandsia transplant survival. A ) Transplants were first immers ed in a buc ket of fresh water , B) gently s haken to remove excess water , and then visually scored for the percent of transplant e xhibiting C) dead, brown tissue and D) live, green tissue .


61 Figure 3 3. Live oak and experimental shade mimic effects on light. Light l evel, expressed as Photosynthetically Active Radiation, in Bahia savannas (grass) and A) beneath adjacent oaks and B) in experimental c ontrol light and shade plots . Data are mean ± standard error.


62 Figure 3 4 . Design and environmental effects of exper imental mesocosms. Mesocosms were stocked with either A) oak limb sections only or B) Tillandsia festoons draped over oak limb sections used to test whether Tillandsia influences invertebrate (isopod) survivorship within oaks in the absence of spider preda t ors. Mean daytime temperature C) and mean daytime humidity D ) in oak only (black) and oak + Tillandsia (grey) treatment mesocosms over the 7 day experiment. Data are shown as the mean +/ standard error for 3 replicate mesocosms per treatment.


63 Figure 3 5 Mechanisms of oak facilitation of Tillandsia . Summary of two experiments testing the presence and mechanisms underlying oak facilitation of Tillandsia . Effect of oak presence (a), meaning Tillandsia festoons were positioned on limbs (oak limb control and procedural control), and oak limb removal, meaning festoons were positioned on the ground, on survival at two sites. Effect of shade and elevation treatments on Tillandsia transplant mean daytime temperature (b) and survival (c), shown on a log scale, over the 2010 study period. Data are shown as mean +/ SE in a and c. Arrow in c denotes when transplants reached 0% survival.


64 Figure 3 6 . Tillandsia effects on canopy temperature and humidity. D ifference in A) mean temperature , B )relative humidity , and C) the coefficient of var iation in relative humidity between Tillandsia colonized and un colonized limbs in two oaks. Points positioned below the dashed zero line denote days where mean temperature or humidity levels were lower and humidity more stable (i.e., less variable) on Tillandsia colonized limbs relative to paired un colonized limbs. Data summarized as the mean difference +/ SE in daytime mean for 3 replicate iButton pairs per oak.


65 Figure 3 7 Oak Tillandsia facilitation cascades across lat itude. Oak Tillandsia facilitation cascades across latitude. Invertebrate A) total density , B) juvenile density , C) species richness , D) Shannon , E) number of feeding guilds , , G) rarefied species richness , and H) rar efied fee ding guild richness on Tillandsia colonized (black) vs. un colonized (white) 1 m 3 live oak limb plots at 6 sites listed by state on the x axis. Data are shown as mean +/ SE. Note log scale in A and B . Adjacent table reports results from ANOVA and randomi zation tests.


66 Figure 3 8 . The Foundation Species Biodiversity model. A ) The Foundatio n Species Biodiversity model and B ) visual diagram of the three t ypes of facilitation cascades it predicts. Where secondary foundation species increase habitat avai lability within a primary foundation species, we predict they enhance abundance (Type A); where secondary foundation species exhibit functional traits different from those of the primary foundation species, we predict they enhance diversity (Type B), and w here they both increase habitat availability and exhibit different functional traits, they enhance abundance and diversity (Type C). Abundance and diversity increases are relative to communities associated with a primary foundation species alone. Primary f oundation species (tree) may directly facilitate communities, as shown in the left panel, or interact with secondary foundation species (fern) to indirectly facilitate communities in a Type A, B, or C facilitation cascade (middle and right panels). The num ber of fern icons represents the relative increase in habitat availability generated by the secondary foundation species (many ferns= large increase, one fern= little increase), and the color of fern icons denote whether they exhibit similar (black like th e tree) or different (green) functional traits than the primary. Invertebrate icons refer to feeding guilds; node size represents abundance; and lines connecting nodes indicate species interactions.


67 CHAPTER 4 SECONDARY FOUNDA T I ON SPECIES GENERATE HOTSPO TS OF MULTIFUNCTIONALITY IN A COASTAL ECOSYSTEM Introduction Biodiversity can be a key driver of ecosystem functioning and where ecologically diverse organisms co occur , multiple functions, such as net primary production, decomposition and water flow, may be maximized simultaneously (Hector and Bagchi 2007 , Gamfeldt et al. 2008) . While decades of research has focused on how man made or experimental gradients in diversity affect ecosystem functions ( e.g., Zavaleta et al. 2010, Hooper et al. 2012, Pasari et al. 2013) , relatively little work has focused on the drivers of natural variation in diversity and how these natural gr adients in diversity then affect ecosystem functioning. Along with physical factors, such as disturbance (Connell 1978, Sousa 1984, Hughes et al. 2007) and resource availability (Bobbink et al. 2010) , species interactions can also shape diversity patterns within landscapes, for instance by ame liorating environmental stress. Such facilitative interactions are perhaps most notable where foundation species ( sensu Dayton 1972) , such as forest forming conifers and reef building corals, provide shelter for many other organisms (Bruno and Bertness 2001, Brooker e t al. 2008, Silliman et al. 2011) . Secondary (dependent) foundation species can be among the organisms facilitated by foundation species and can further enhance diversity through the creation of more, complex habitat (Yakovis et al. 2008b, Bishop et al. 2012) . A posi tive direct interaction among foundation species that gives (Altieri et al. 2007) and been documented in many ecosystems (Angelini et al. 2011, Thomsen et al. 2013) . Despite the links between facilitation cascades and diversity and, in turn, between diversity and ecosystem functioning, ecologists have yet to integrate these


68 complementary fields an d test whether secondary foundation species, as the key intermediate link in facilitation cascades, augment diversity and multiple ecosystem functions at either the patch or larger landscape scale. Within landscapes (scale: 100s of m 2 ) structured by prim ary (dominant) 1m 2 ) is predicted to be positively correlated with ecological diversity ( The Foundation Species Biodiversity hypothesis, Angelini and Silliman 2014) . While it is intuitive that as a secondary foundation species provides more habitat structure it will support a higher number and diversity o f organisms, few studies have conducted gradient experiments to density and the diversity of associated species (Bishop et al. 2012, Byers et al. 2012) . Distinguishing whether habitat generation by a secondary foundation species elicits a linear, saturating, or e xponential increase in diversity, for example, can provide insight to the mechanisms that control the settlement of other organisms (e.g. habitat availability, intraspecific competition or facilitation) and help pinpoint which patches are most likely to su pport high diversity within and across landscapes. Through their own activities and those of the organisms they support, secondary foundation species are also almost certain to modify multiple ecosystem functions (Angelini and Silliman 2014) . Yet, how density of secondary foundation species regulates individual ecosystem functions and multifunctionality the simultaneous consideration of multiple functions (Gamfeldt et al. 2008, Byrnes et al. 2014) within a patch is unknown. Finally, the effects of secondary foundation species at larger spatial scales larger than a patch has not been measured a step tha t is necessary to test if variation in the distribution of


69 secondary foundation species can drive measureable differences in diversity and ecosystem functioning at landscape levels. Improving our understanding of how much secondary foundation species regul ate diversity and ecosystem functioning on larger spatial scales will provide much needed perspective on the ecological significance of facilitation cascades, a relatively new concept (Altieri et al. 2007) , and insight to the potential efficacy of conservation strategies that integrate secondary foundation species into management plans (Angelini et al. 2011) . Salt marshes are highly productive grasslands that form in temperate, wave protected, intertidal environments and provide a useful system for investigating the links between secondary foundation species, diversity, and ecosystem functioning. In the southeastern Atlantic coast of United Stat es , Spartina alterniflora (smooth cordgrass, her e after cordgrass ) generates much of the three dimensional structure of salt marshes and plays a key role in sustaining a number of the services that this ecosystem is valued for including shoreline protection , nutrient filtration, and nursery habitat provision (Barbier et al. 2011) . Embedded in the mud around cordgrass stems, the ribbed mussel, Geukensia demissa (hereafter mussels) occurs as solitary individuals a nd in clumped aggregations of up to ~80 individuals in higher elevation marsh platforms. On these expansive marsh platforms that comprise up to ~30% of the total area of salt marsh (Schalles et al. 2013) , the surface is flooded only briefly by the tides and the temperature of exposed mud can exce ed 46°C at the marsh surface (C. Angelini, unpublished data ), levels well above the thermal limit of mussels (Jost and Helmuth 2007, Altieri et al. 2007 , Appendix A ) . Experimental field studies during summer months have shown that long term mussel survival on these marsh platforms is dependent on


70 canopy, as nearly all mussels die in the absence of marsh grass stems ( Altieri et al. 2007 , Appendix A ) Within cordgrass monocultures, our observations indicate that mussels function as secondary foundation species as they both are dependent on a facilita tion by a primary foundation species ( Altieri et al. 2007 , Appendix A ) and independently facilitate a number of resident macro invertebrates (i.e. those that do not migrate with the tides), including a diversity of burrowing crabs (CA and JNG, personal obs ervation ). In this study, we experimentally created replicate mussel aggregations spanning t he natural density range of aggregation size observed in marsh platforms and quantified responses of five invertebrate functional groups, seven ecosystem functions , and two distinct multi functionality indices (Gamfeld t et al. 2008, Byrnes et al. 2014) . We hypothesized that increasing mussel density would generally increase diversity, individual ecosystem functions, and multifunctionality. Furthermore, we hypothesized that diversity would be positively related to multi functionality due to differences among invertebrate functional groups in activities performed. To gain insight to the mechanisms by which mussel density may control diversity and ecosystem functioning, we used model selection techniques to identify the fun ctional form of each relationship (Cardinale et al. 2011) . Finally, we used information from mussel aggregation surveys and our patch scale experiment to begin to estimate how much mussels af fect invertebrate diversity and salt march ecosystem functioning at larger landscape scales. M ethods Study System W e performed this research within the Sapelo Island National Estuarine Research Reserve on Sapelo Island, G eorgia (31°24'26 "N , 81°17'24 "W ), a barrier


71 island embedded an expansive network of salt marshes that spans the mouth of the Altamaha River estuary. On Sapelo Island, we conducted the field experiment and mussel distribution survey s i n marsh platforms, relatively flat expanses of salt marsh that are dominated by cordgrass monocultures whose canopy reaches ~ 50cm in height by the end of the growing season in early fall ( Pennings et al. 2005, Schalles et al. 2013) . In these marsh platforms, mussels exhibit a widespread distribution and occur in aggregations that vary in mussel density, from 1 to ~100 individuals. The distribution of cordgrass, mus sels, and resident macro invertebrates at our field sites are typical of marsh platforms throughout the southeastern US (CA, unpublished data ). Density Dependent Effects of Mussels on Biodiversity and M u ltiple Ecosystem Functions: an E xperiment Prior to se tting up our experiment to test the effects of mussel aggregate density on invertebrate diversity and ecosystem functioning, we first needed to characterize the range of densities with which mussels naturally aggregate in marsh platforms. To do so, we haph azardly identified 17 mussel aggregations that captured the range of sizes observed within 50m × 50m areas at each of two marsh platform sites (hereafter, sites 1 and 2), extracted each aggregation, and then counted and measured every mussel retained in a 0 .5cm sieve over which aggregations were washed. Natural aggregations contained between 1 34 and 1 82 individual mussels, which were 5.7 ± 2.2 and 7.2 ± 2.9cm (mean ± standard deviation) in length at sites 1 and 2, respectively (see Appendix B for addition al data). At site 1, we then marked 24 plots, positioned >1.5m apart, in a monoculture of cordgrass. Plots were cleared of resident mussels in April 2012 and randomly assigned one mussel density treatment ( 0, 1, 3, 5, 10, 20, 40, or 80 mussels , with densi ty levels


72 based on our natural range of mussels per aggregation, n= 3 replicates per treatment). We then transplanted the appropriate number of mussels ( length: 50 80mm) collected from a nearby marsh platform in a cluster to mimic natural aggregations in e ach plot. Three dead mussels were replaced on day seven of the experiment , after which aggregations were left undisturbed until the following summer . In August 2013, we used the following methods to quantify the effect of mussel density on five invertebrat e functional groups, seven ecosystem functions, and two measures that integrate the responses of all seven functions the average multifunctionality index and threshold index (Byrnes et al. 2014) in 0.5m × 0.5 m plots centered on each experimental treatment. The plot area encompassed all 80 mussels transplanted in our largest aggregations. Invertebrate Functional G roup D iversity To assess the effect of mussel density on diversity, we counted every macro inverteb rate in each plot. We classified invertebrates using five functional groups (mud crabs, marsh crabs, adult fiddler crabs, juvenile fiddler crabs, and snails) rather than taxonomic species to account both for functional redundancy between species (e.g. sinc e Eurythium limosum and Panopeus obesus are generalist predators that excavate similarly sized burrows, we counted both as mud crabs) and functional disparity between life stages of the same species (e.g. since adult fiddler crabs excavate burrows that are 10× wider and deeper than those excavated by juvenile fiddler crabs, we counted them separately). At this field site, burrow densities correspond closely to crab densities [0.95 mud crabs per burrow JNG unpublished data , 1.2 marsh crabs ( Sesarma reticula tum ) per burrow S. van Montfrans unpublished data , 0.85 adult and 0.94 juvenile mud fiddler crabs ( Uca pugnax ) per burrow, CA unpublished data ) , so we


73 counted burrows as a non destructive measure of each crab functional group. Snails were counted on the m arsh surface and cordgrass canopy. From the invertebrate counts, we calculated functional group richness and the effective number of functional groups, D (MacArthur 1965) (Jost 2006) . We report D , linear scales, because it provides a simple, interpretable, and general measure of diversity: if D =2 in plot A and D =3 in plot B, for instance, we can conclude that there are 2 and 3 equally common functional groups in plots A and B, respectively, and that plot B is 1.5 times more diverse than plot A (Jost 2006) . In vertebrate b iomass We assessed invertebrate biomass, a measure of secondary marsh productivity, in each plot by collecting a random sample of 20 mud crabs, 20 marsh crabs, 20 adult and 20 juvenile fiddler crabs, and 20 snails at the site of our exper iment. Each invertebrate was dried in a 60°C oven for 48 hours and weighed. We then calculated invertebrate biomass by multiplying the density of each functional group by the respective average biomass per individual, and summing these values across all fu nctional groups in each plot. Soil accretion We measured soil accretion , the vertical accumulation of settled, but not yet root bound, soil , by inserting a 3mm diameter rod perpendicularly into t he marsh until it made contact with the rigid root mat (Smith and Frey 1985) . We recorded the unbound soil depth at five haphazardly chosen locations per plot in October 2013 and averaged these values to generate integrative measure of the extent to which soil accreted on the


74 marsh surface over the duration of the experimen t. Additional data on the short term soil deposition rate is pro vided in Appendix C . Decomposition We quantified decomposition, a key process in nutrient cycling, using bait lamina test s (Terra Protecta, Berlin, Germany). In August 2013, we haphazardly inserted 3 bait strips to a depth of 12cm in each plot, collected them after 48 hours, and counted the number of baits that were decomposed out of the 16 baits per strip (Simpson et al. 2012) . Supplemental methods used to measure decomposition and results are provided in Appendix D . Infiltration rate W e measured infiltration, the rate with which water percolates through marsh soils, by securing a 12cm diameter double ring infiltrometer to the marsh surface, filling it with two liters of creek water four hours after high tide , and recording the time required for the water to drain (Hensel and Silliman 2013) . Infiltration is a critical function of salt marshes that prevents the development of water logged, anoxic soils which limit primary and secondary production (Mendelssohn and Sen eca 1980) and promotes the uptake and filtration of nutrients from terrestrial and estuarine water sources (Hemond and Nuttle 1984, Harvey and Nuttle 1995) . Benthic algae biomass W e quantified the density of benthic algae , a major component of salt marsh primary production and a dominant resource consumed by fiddler crabs and snails (Miller et al. 1996) , on the surface of the marsh using a hand held fluorometer ( Bentho torch , bbe Moldaenke Gmb H, Germany). On each of three consecuti ve sunny days, we


75 recorded three readings ( µ g diatoms + µ g cyanobacteria per cm 2 ) per plot, which we then averaged to derive one integrated measure of benthic algae biomass . Aboveground cordgrass biomass W e harvested , rinsed, dried in a 60°C oven for 48 hours, and weighed all live stems located within plot boundaries in October 2013 to quantify aboveground cordgrass biomass . Standing plant biomass, is commonly used as a proxy for primary production and is a function t hat mediates carbon sequestration and wave attenuation, two of the key ecosystem services provided by salt marshes (Barbier et al. 2011) . Belowground cordgrass biomass We used a 4.5 cm diameter corer to extra ct 5 replicate belowground biomass samples from each plot. Replicate cores within plots were pooled before roots and rhizomes were rinsed over 0.5mm sieves, collected, dried in a 60°C oven for 48 hours, and weighed. We assayed belowground cordgrass biomass because this function also contributes to regulating carbon sequestration in salt marshes. Multifunctionality To distinguish whether mussel density broadly enhances multiple ecosystem functions to increase multifunctionality or possibly increases some functions at the cost of others, we calculated the average multifunctionality and multiple threshold indices described in Byrnes et al. 2014 . To calculate t he former, we standardized each function to the same scale by dividing the value of each function measured in each plot by the average of the four highest values measured for that function across all plots and then averaged together all seven standardized function values for each plot. We assume that the high values of each of our functions indicate a high level of functioning; for example, high soil accretion


76 denotes a high level of performance for this function. The average multifunctionality index can be interpreted as the average level of all seven functions. Because one cannot interpret whether all functions are being performed simultaneously at a high level as functions that are performed at low lev at high levels using this index, we also tallied the number of functions, at their standardized function value, in each plot that surpassed each of five threshold levels: 10, 30, 50, 70, and 90% of maximum funct ioning. The resulting, multiple threshold index scores, which range from 0 to all 7 functions, can be interpreted simply as the number of functions performed above a given threshold level in a plot (Zavaleta et al. 2010, Byrnes et al. 2014) . Analyses Using generalized linear models (glm) and non linear least squares (nls) models, we fit null ( Y i = a), linear (Y i = a + bM), log [Y i = a + b*log(M+1)], power ( Y i = a + cM z ) , and hyperbolic [Y i = aM/(b+M)] relationships between the number of mussels added (M) and each functional group and ecosystem function response variable (Y i ) and selected the best fitting model using size, AICc (Burnham and Anderson 2002) . We investigated each of these models because the mechanism by which mussels most strongly influence a specific response variable fit model: a linear model suggests a response variable is controlled by mussel density, a log model by an effect of mussel density that diminishes as aggregations get larger, a power model by processes that scale proportionally with mussel density, such as those that may be controlled by the area of an aggregation which scales to the z= 0.6 power with mussel d ensity (see below), and a hyperbolic model by the maximum carrying capacity of a


77 plot. Model fits, AICc values, and AICc weights for invertebrate and ecosystem function response variables are reported in Appendices 5 and 6, respectively. For response varia bles in which the best fit model was linear or log, we report the significance of Mussel Treatment as the probability (P) of obtaining the slope value, b, given that the null hypothesis (that b equals zero) is true. For response variables in which the best fit model is a power or hyperbolic, we report the significance of Mussel Treatment as the probability (P) of obtaining the slope (c) and exponent (z), or asymptote (d) and half maximum value (k), value given that the null hypothesis (that each parameter e quals zero) is true. Analyses were conducted in R version 3.0.2 (R Core Development Team 2012) and model comparisons were conducted using the AICcmodavg package (Mazerolle 2013) . Effect of M ussels on B iodiversity , in turn, on M ultifunctionality To summarize these results, we inv estigated the functional form and significance of the relationships between mussel density and the effective number of functional groups, D , and between D and average multifunctionality index using the model selection approach just described. The effect of D on average multifunctionality must be interpreted with caution however: since D is not independent of mussel density, we cannot discern the relative importance of D versus mussels in driving the observed correlation. Nevertheless, we conducted these ana lyses to explore whether mussels might indirectly regulate multifunctionality through their direct effects on invertebrate diversity and how closely variation in multifunctionality was related to differences in diversity a cross our experimental plots.


78 Land scape Level Effects of Mussel Aggregations on B iodivers ity and Multifunctionality: a S urvey To gauge whether mussels enhance biodiversity and ecosystem functioning at larger spatial scales, we extrapolated our experimental, patch scale results to the marsh landscape. Specifically, we used a GPS (Trimble R6 GNSS, Sunnyvale, CA) to map and measure the area of every mussel aggregation found within a 10m × 50m marsh platform plot at sites 1 and 2 (see Study System ) . We then estimated the number of mussels (M) i n each aggregation using the following equations: M = 70.34 × A 0.67 (site 1) and M= 86.89 × A 0.53 (site 2), where M is the number of mussels and A is the aggregation area in m 2 (see Appendix B ). Next, we divided each 500 m 2 surveyed area into 2000, 0.5m × We then used the best fit model equations from our experiment (Table 4 1) to estimate the density of e ach invertebrate functional group and level of each ecosystem function in each cell and summed each invertebrate and function metric across all mussel occupied cells (X i, Mussels ) and unoccupied (X i, No Mussels ) cell values. Finally, we estimated the contri bution of mussel occupied cells (Z i ) to supporting the total number of each invertebrate functional group and the total production of each ecosystem function across the surveyed area as: Z i = 100 × [ X i, Mussels / (X i, No M ussels + X i, Mussels ) ] . R esults As m ussel aggregations increased in size, the diversity of resident invertebrates and abundance of three of the five functional groups inc reased at the patch scale (Figure 4 1 A G ). Specifically, mud crab density and functional group richness both increased as a log function, marsh crab density as a power (z= 0.83) function, and


79 juvenile fiddler crab density as a linear function of mussel density (Table 4 1). Mud, marsh, and juvenile fiddler crabs were absent or rare in 0 mussel plots and highly concentrated in large mussel aggregations. In contrast, adult fiddler crabs were more abundant in plots with 0 or few mussels than in larger aggregations and snails were distributed evenly across our experimental plots. Along with the pronounced variation among experime ntal treatments in invertebrate composition and density, five of the seven ecosystem functions also varied significantly as a function of mussel density. While, invertebrate biomass, decomposition and water infiltration increased as a linear function of mu ssel density , soil accretion and aboveground Spartina biomass increased as a log function of mussel density (Figure 4 1, Table 4 1) . B elowground Spartina biomass and benthic algae biomass, in contrast, did not vary significantly across experimental plots. Average multifunctionality, an integrative measure of all seven ecosystem functions, nearly doubled from 0 mussel to 80 mussel plot s (44.5 versus 81% of maximum functioning, respectively, Figure 4 1N ) and increased as a power function (z=0.64) with increas ing mussel density . Lik ewise, as mussel density increased, so did the number of function s being performed above 10, 30, 50, 70, and 90% of maximum functioning thresholds . Threshold level af fect ed the shape of the relationship between the number of mussels added and the number of functions performed above or equal to the threshold: while the number of functions performed above most thresholds (10, 50, and 70%) increased as a linear function of mussel density, it increased as a log function at the 30% thresho ld and power function (z=2.62) func tion at the 90% threshold (Figure 4 1O, see Appendix E for model compariso ns and best fit model equations ) .


80 In addition, we found that mussel density drove a significant linear increase in t he effective number of functio nal groups D , a measure of diversity (R 2 =0.72, P<0.0001, Figure 4 2 , Table 4 1, Appendix E ). In turn, D was significantly, positively, and linearly correlated with average multifunctionality (R 2 =0.60, P <0.0001, Figure 4 2G ). At sites 1 and 2, we observe d 111 and 142 mussel aggregations, which collectively occupied 6.3 and 4.3m 2 , or 0.8% and 1.3%, of the 500 m 2 of salt mars h platforms surveyed, respectively. Based on the total number and size fre quency distribution (Appendix F ) of mussel aggregations, we estimate that mussel aggregations at sites 1 and 2, respectively support 100% and 100% of mud crabs, 100% and 100% of marsh crabs, and 11% and 19% of juvenile fiddler crabs, but less than 7% of large fiddler crabs and snails that inhabit the marsh platform (Figure 4 3 ). Likewise, we estimate that mussel aggregations produced 23% and 30% of overall soil accretion and 98% and 99% of the infiltration function ing at the landscape scale (Figure 4 3 ). In contrast to their disproportionately large contribution to soil accretion and infiltration decomposition, invertebrate biomass, above and belowground Spartina biomass, and benthic algae biomass f unctions on a landscape scale. Discussion Ou r field experiment revealed that as mussel aggregations increase in size, they ecosystem functions can be linked to the activities of the resident invertebrates that sig nificantly increase in density in the presence of mussels, our data suggest that a key mechanism by which secondary foundation species increase multifunctionality as they increase in density is through their facilitation of a more abundant and diverse


81 asse mblage of associated species. In addition to enhancing diversity and ecosystem functioning at the patch scale, we found that although mussel aggregations occupy a small fraction of the marsh platform landscape, they support much of the resident crab predat or population, maintain key water drainage channels, and contribute substantially to the vertical accretion of salt marshes through their extraction and deposition of suspended sediment. Collectively, these findings indicate that the abundance and density distribution of secondary foundation species can be important, but currently underappreciated, drivers of variation in diversity and functioning within and among landscapes. Densi ty D ependent E ffects of Mussels on Invertebrate Diversity and Ecosystem F unc tions The results of our experiment are consistent with other facilitation studies that document pronounced changes in ecological diversity where primary foundation species facilitate secondary foundation species (Altieri et al. 2007, Yakovis et al. 2008b, Bishop et al. 2012, Angelini and Silliman 2014) . In our salt marsh system, mussels are effective secondary foundation species that enhance cordgrass habitat complexity in three obvious ways: their shells which provide hard substrate, their pseudofeces which create a soft layer of sediment (Smith and Frey 1985) , and their deposition of nutrients that stimulates cordgrass growth (Bertness 1984) . As mussel aggregations increase in size, we observed concomitant increases in the number of mud crabs that use shells for burrow support and protection (Silliman et al. 2004) , juvenile fiddler crabs that appear dependent o n soft pseudofeces layer where strong appendages are not required to excavate burrows, and marsh crabs that tend to concentrate within lush patches of cordgrass where the mud remains cool and moist (MJS Hensel, personal


82 communication ). Interestingly, mud, juvenile fiddler, and marsh crab density increased as log, linear, and power functions of mussel density, respectively, suggesting different mechanisms control the settlement and retention of these ecologically distinct functional groups at this scale (Sillim an et al. 2004) . In contrast to these crabs that associate strongly with mussels, adult fiddler crabs occurred at higher densities in plots with 0 mussels, indicating the benefits this species gains by inhabiting mussel aggregations at juvenile life stag es become offset by costs, such as higher mud crab predation risk, as they mature. Snails, a species that may balance competition with adult fiddler crabs for benthic algae food outside of aggregations (Currin et al. 1995) with risk of predation by mud and marsh crabs on aggregations (Silliman et al. 2004, Soomdat et al. 2014) , were similarly abundant throughout the marsh platform. In the absence of experiments te sting the relative importance of food availability, predation, and other factors, the specific mechanisms underpinning the diverse responses of different invertebrate functional groups to increasing mussel density we observed remain speculative and a topic for further study. Regardless of the precise drivers, invertebrate communities were less abundant and diverse in plots structured by cordgrass alone than by both c ordgrass and many mussels (Figure 4 1A F , Figure 4 2 ). Coincident with the increasing conce ntration of invertebrates, five ecosystem functions and multifunctionality also increased as the size of muss el aggregations increased . Invertebrate biomass, d ecomposition , soil accretion, infiltration, and aboveground cordgrass biomass functions achieved maximum levels on the largest, 80 mussel aggregations, responses likely to be driven at least in part by: improved invertebrate growth and survival due to enhanced bottom up (e.g. food quality


83 and/or quantity) and reduced top down control (e.g. blue crab, redfish, and raccoon predation) mediated by higher mussel habitat complexity (Currin et al. 1995, Silliman et al. 2004) , enhanced oxygen penetration due to the proliferation of juvenile fiddler crab burrows (Bertness 1984) , increased sediment accumulation due to persistent mussel deposition of pseudofeces (Smith and Frey 1985) , expanded channels for water flow due to mud and marsh crab excavation of deep, wide burrows (Hensel and Silliman 2013) , and elevated nutrient availability due to invertebrate excretion of waste and stimulation of microbial activity (Bertness 1984, 1985) , respectively. Similar to the response of invertebrate functional groups, some ecosystem functions increased as a linear function, others as a log function, and average multifunctionality as a power function of mussel density (Table 4 1, Figure 4 1). This result suggests mussels, within the range of densities we tested, stimulate different ecosystem functions by different mechanisms. Soil accretion, for example, may have increased as a log function of mussel density because individual mussels produce relat ively less pseudofeces as aggregations grow larger due to increased competition for filtrate, while invertebrate biomass increased as a linear function of mussels because individual mussels add a consistent amount of habitat structure, regardless of aggreg ation size, to support resident invertebrates. Belowground cordgrass biomass and benthic algae biomass, in contrast, did not vary with mussel density. From our methods for measuring these two functions, we cannot determine if the rate of production of root s and benthic algae varied across our plots, but was undetectable due to high decomposition. Our bait lamina tests indicated that decomposition inc reases with mussel density (Figure 4 2H ), however, suggesting


84 that if our results misrepresent how much root material or benthic algae were produced, they likely underestimate these functions more in treatments with many mussels than those with few or none. Bearing in mind these methodological concerns, the positive response of five of the seven ecosystem functio ns and both multifunctionality indices to increasing mussel density reveals that the presence of this secondary foundation species broadly promotes multiple functions through their own activities and facilitation of other, ecologically diverse invertebrate s. Collectively, our field experiment shows that mussels attract and support ecologically diverse communities within landscapes structured by cordgrass, and, in doing so, boost salt marsh multifunctionality on patch scales (Fig ure 4 2). This intuitive fin ding bridges two complementary and active fields in ecology: one focused on whole community, or food web, facilitation by foundation species (Bruno and Bertness 2001, Stachowicz 2001, Bruno et al. 2003, Baiser et al. 2013) and a second on biodiversity effects on ecosystem functioning ( see Cardinale et al. 2012, Hooper et al. 2012 for reviews ) . To further integrate these fields, additiona l research is needed that untangles the independent and interactive effects of foundation species and the ecologically diverse communities they support on ecosystem functioning and explores the generality of our results in other systems. The Importance of Mussels as D rivers of Diversity and M ultifunctionality at Landscape S cales In using mussel distribution surveys to extrapolate our experimental results, we estimate that mussels support the 100% of the mud and marsh crab populations at our survey sites de spite occupying a sma ll proportion of landscape (Figure 4 3). Based on our observation that mud and marsh crabs burrow outside of mussel aggregations in


85 marsh platforms but do so very rarely (i.e. no burrows generated by these functional groups were counte d in our 0 mussel experimental plots), we are certain that this result overestimates the contribution of mussels to supporting these two functional groups. Regardless if the true percentage of mud and marsh crabs that inhabit mussel aggregations is closer to 90 or even 80% in marsh platforms, our results reveal that mussels are significantly altering the overall complexity of salt marsh food webs in providing key habitat for these top (mud crab) and intermediate (marsh crab) resident consumers. Furthermore, because mud crabs are sit and wait, ambush predators that eat fiddler crabs, marsh crabs, and other mud crabs as well as mussels and snails (Silliman et al. 2004) , and marsh crabs forage cryptically at night on small invertebrates and cordgrass tissue (Soomdat et al. 2014) , mussels are likely altering the spa tial structure and temporal dynamics of predator prey interactions in marsh platforms through their facilitation of these particular functional groups. Mussels also support a significant proportion (up to 20%) of the juvenile fiddler crab population, a sp ecies that in adult life stages stimulates cordgrass productivity throughout the marsh by aerating marsh soils and cycling nutrients (Bertness 1985) and accounts for a sizeable fraction of salt marsh secondary production (Montague 1980) . Other studies have also found juvenile organisms concentrated within the complex structures of secondary foundation species (Altieri et al. 2007, Dijkstra et al. 2012, Angelini and Silliman 2014) , hinting that In addition, our spatial analyses imply that mussels contribute substantially to the production of some ecosystem functions and little to oth ers on the landscape scale (Figure 4 3). As a result of their active deposition of silt packed pseudofeces, we


86 estimate that mussels account for up to 30% of t otal marsh soil accretion (Figure 4 3 ), a value that is within the range of estimates for the contribution of mussel pseudofeces to net sedimentation (7 43%) reported at similar marsh elevations in previous studies (Letzch and Frey 1980, Smith and Frey 1985) . Mussel aggregations also appear to be mediating patterns in water drainage in marsh platforms. Specifically, we estimate that more than 97% of infiltration occurs as a result of mussel aggregations, suggesting that mussels mediate how effectively southeastern US salt marsh platforms can drain and filter nutrients. We suspect that much of the effect of mussels on inf iltration is through their facilitation of mud and marsh crabs that excavate channels through which water drains (Hensel and Silliman 2013) . In contrast to the strong influence of mussels on these two functions, they contributed relatively little to the pr oduction of the five other functions when evaluat ed at the landscape scale (Figure 4 3). The discrepancy in the magnitude of mussel effects on these different functions at patch versus landscape scales highlights the importance of integrating multiple ecos ystem functions and multiple spatial scales in ecological studies to create a more complete picture of ecosystem multifunctionality beyond the scale of typical experimental units (Cardinale et al. 2011) . Although it is beyond the scope of this study, we anticipate that our assessment of the relative importance of mussels (and the diversity of organisms they support) will be altered and improved if we were to incorporate other spatial scales in to our analyses: for example, mussels effects may depend on the density and spatial pattern of other aggregations at intermediate scales (several m 2 ) and on marsh elevation or proximity to sources of mussel larvae at whole estuary scales (Kuenzler 1961, Bertness and Grosholz 1985, Stiven and Gardner 1992, Pringle et al. 2010) .


87 Finally, our field survey revealed that the density and aggregation size distribution of mussels diff ered in the two marshes (Appendix F ), a finding that was not in itself surprising. As a result of this natural variation in mussel distribution, however, we estimate that these two marshes differed notably in the number of invertebrates they support and pe rformance of func tions they maintain (Figure 4 3 ). Although our approach to estimating landscape diversity and functioning from patch scale results may be too simplistic to predict the total abundance of invertebrates or level of functioning over the lands cape precisely, our results hint that ecosystems structured by a moderate to high cover of secondary foundation species may be far more effective in sustaining biodiversity and multiple functions than those with low cover or none. Integrating Secondary Fo undation Species into Natural Resource M anagement After several decades of theoretical and empirical advances, the field of biodiversity ecosystem function theory has begun to significantly influence on environmental policy and conservation. In particular, natural resource management strategies long focused on preserving biodiversity are increasingly considering the potential benefits of biodiversity for maintaining a wide range and high level of ecosystem services (Tscharntke and Klein 2005, Cardinale et al. 2012) . Among the greatest challenges facing these management efforts is identifying how best to protect biodiversity in an economically and ecologically sustainable way. Our findings shed light o n this issue as they reveal that facilitation by foundation species can promote biodiversity and its corresponding effects on ecosystem functioning at patch scales and drive striking spatial heterogeneity in these variables across landscapes. Consequently, strategies aimed at fostering the growth of primary foundation species and coverage of


88 secondary foundation species are likely to be quite effective in augmenting biodiversity and functioning over the long term and at relevant (large) spatial scales.


89 T able 4 1. Summary of best fit models predicting the relationship between the number of mussels in an aggregation (M) and number of each invertebrate functional group or level of each ecosystem function (Y). Mussel treatment significance Best fit model T slope T exponent P slope P exponent Mussel density = M Invertebrate response Functional group richness Y = 0.69*log(M+1) + 3.35 7.99 NA <0.0001 NA Mud crabs Y = 0.33*log(M+1) 7.13 NA <0.0001 NA Marsh crabs Y = 0.074M 0.83 0.64 2.19 0.0394 0.527 2 Adult f iddler crabs >5mm Y = 0.76*log(M+1) + 8.73 2.23 NA 0.0364 NA Juvenile f iddler crabs <5mm Y = 1.15M + 5.29 8.77 NA <0.0001 NA Periwinkle snails Y = 0.69M + 197.7 1.84 NA 0.0791 NA e Diversity Y= 0.011x + 1.48 3.66 NA 0.0014 NA Ecosystem function response Soil Accretion Y = 1.29*log(M+1) + 0.49 11.51 NA <0.0001 NA Decomposition Y = 0.29M + 55.23 2.87 NA 0.0089 NA Infiltration rate Y = 0.79M + 0.0067 6.87 NA <0.0001 NA Aboveground biomass Y = 10.7*log(M+1) + 44.81 3.21 NA 0.0042 NA Bel owground biomass Y = 15.48 NA NA NA NA Benthic algae biomass Y = 16.56 NA NA NA NA Invertebrate biomass Y = 0.46M + 11.97 5.89 NA <0.0001 NA Multifunctionality indices Average Multifunctionality Y = 2.26M 0.64 + 40.91 1.91 5.56 0.0706 <0.0001 # f unctions >10% threshold Y = 0.013M + 6.04 5.06 NA <0.0001 NA # functions >30% threshold Y = 0.76*log(M+1) + 3.30 5.84 NA <0.0001 NA # functions >50% threshold Y = 0.049M + 3.08 7.52 NA <0.0001 NA # functions >70% threshold Y = 0.048M + 2.00 6.42 NA <0.0001 NA # functions >90% threshold Y = 0.052M 2.62 + 0 .00004 0.21 2.35 0.8397 0.0286


90 Figure 4 1. Effect of mussel addition on invertebrate diversity and multifunctionality. T he density dep endent effect of mussels on: A) mud crabs , B) marsh crabs , C) adult mud fiddler c rabs , D) juvenile mud fiddler crabs , E) snails , F) functional group richness , G) invertebrate biomass , H) soil accretion , I) decomposition , J) infiltration rate , K) a boveground cordgrass biomass , L) b elowground cordgrass biomass , M) benthic algae biomass , and two measures of salt marsh multifunctionalit y: N) average multifunctionality and O) the number of functions exceedin g a range of threshold values . Invertebrate response variables are s hown as the number per plot (A F ) and units of measur ement for ecosystem functions (G O ) are indicated in smaller font below panel titles. Points and error bars denote the mean and standard error of 3 replicate plots per mussel addition treatment. Fitted lines in (A O ) reflect the linear or non line ar model that be st fit each response variable (A N) and each threshold level (O ). Equations for best fit lines are in Table 4 1.


91 Figure 4 2 Effect of mussels on diversity and diversity on multifunctionality. Resident salt marsh functional groups inclu de: A) ribbed mussels, B) mud crabs, C) marsh crabs, D) fiddler crabs, and E) snails. The effect of F) mussel density in experimental aggregations on functional diversity, shown as the effective number of functional groups, D , and G) the corresponding effe ct of D on multifunctionality, shown as the average % of maximum functioning for the seven functions quantified in our experiment.


92 Figure 4 3 . Landscape effects of mussels on marsh diversity and multifunctionality. A) A cordgrass dominated southeast ern US salt marsh platform in which aggregations of ribbed mussels create hotspots of biodiversi ty and ecosystem functioning . B) Map of the distribution of mussel aggregations (1, 2, 3, 4 20, 21 40, and 40 82 mussels per aggregate in different colors) at Site 1 and estimated contribution of mussel aggregations at their natural distribution to supporting C) the cumulative number of resident marsh invertebrates and D) performance of seven marsh ecosystem functions (d) in 500 m 2 areas at two marsh platform s ites surveyed on Sapelo Island, GA.


93 CHAPTER 5 REMNANT PATCHES, A KEYSTONE MUTUALISM, AND THE RESILIENCE OF A COASTAL ECOSYSTEM TO DROUGHT Introduction Episodes of climate stress, such as drought and heat spells, are increasing in frequency and severity (IPCC 2014) , eliciting alarming changes in ecosystem structure and f unction worldwide (Folke et al. 2004, Lamb et al. 2005, Silliman et al. 2013) . While some ecosystems exhibit remarkable resilience to these perturbations and return to pre disturbed states swiftly, others recover slowly or even pers ist in disturbed states (Pascual and Guichard 2005, Lotze et al. 2006) . On e factor that contributes to this observed variation in resilience is the reproductive mode of the foundation species, the organism that largely defines communities (Dayton 1972, Bertness and Callaway 1994, Bruno et al. 2003, Elli son et al. 2005) . Relative to foundation species that re sprout from rhizomes ( tundra tussocks: Syndonia Bret Harte et al. 2013) , g erminate from seed banks ( giant kelp; Carney et al. 2013) ), or disperse via seeds or fragments ( braching corals; Smith and Hughes 1999) , those that expand incrementally via clonal or clone like growth are predicted to recover sluggishly, particularly when disturbances are large (Bullock et al. 1995) . Classic studies of plant succession and recent research on forest restoration have shown that patches of mature itions from one state to another through their clone like growth (Yarranton and Morris on 1974, Franks 2003, Zahawi and Holl 2013) . Here we explore the importance of


94 the number and spatial distribution of remnant patches that survive episodes of climate stress me diate ecosystem recovery. In clonal growth dependent ecosystems, foundation species may recolonize disturbed areas from two potential sources: individuals surviving along the perimeter of the disturbance and in remnant patches (Angelini and Silliman 2012, Zahawi and Holl 2013) . Because perimeter: area ratios increase as perimeter should be a relatively more important source for recovery when disturbances are small than large and contribute progressively less to re colonization over time as disturbance areas close (Paine and Levin 1981, Sousa 1984, 2001) . Like wise, because patches grow radially within disturbances, they should be a relatively more important source of re colonization when disturbances are large than small and contribute progressively more to recovery over time (Zahawi and Holl 2013) . Based on these geometric relationships, patches have th e potential to accelerate ecosystem transitions from damaged to healthy states (e.g. Figure 5 1A ); whether they do so in reality is unknown. To examine the influence of patches in the recovery dynamics of real ecosystems, studies are needed that quantify t he degree to which remnant patches shorten ecosystem recovery intervals (relative to disturbed areas with no patches) at the density and spatial distribution with which they naturally occur and rate at which they naturally expand. If these efforts reveal t hat remnant patches control ecosystem recovery, the logical, next step will be to elucidate the mechanisms that facilitate patch survival during climate stress (i.e. increase resistance) to identify how patch dynamics initially arise.


95 Along the Gulf and s outheastern Atlantic coasts of the US, salt marshes dominate wave protected, intertidal environments and offer a relevant ecosystem for testing whether patches enhance ecosystem recovery. Over the past decade, severe and frequent drought has been associate d with mass die off of smooth cordgrass ( Spartina alterniflora , hereafter cordgrass) , the marsh forming foundation species in this region that expands nearly exclusively via clonal growth (Travis and Hester 2005, Angelini and Silliman 2012 , Appendix G for experiment al confirmation of cordgrass' dependence on clonal growth ) . During drought, typically saturated marsh soils can dry out, causing salt, acid, and heavy metals to concentrate around cordgrass roots and snail grazing to increase on live grass leaves (McKee et al. 2004, Silliman et al. 2005, Palomo et al. 2013 , Figure 5 1B ) . The denuded mudflats that result from this lethal cockta il of physical and biotic stress have been so expansive and widespread (McKee et al. 2004, Alber et al. 2008) that it seems possible that they may be persistent features in many of these ecologically and economically val uable ecosystems (Barbier et al. 2011) . At the end of a severe drought in late 2010, we observed remnant cordgrass patches (i.e. clusters of live stems between ~0. 01 1.5m 2 ) surviving in variable densities and spatial distributions within recently formed mudflats on Sapelo Island, Georgia ( Figure 5 1B,C ). Upon closer inspection, we noted that many of the remnant cordgrass patches were associated with ribbed mussels ( Geukensia demissa , hereafter mussels) , hinting that the presence of intertidal marsh environments and reciprocally stimulate cordgrass growth through nutrient deposition (Bertness 1984, Altieri et al. 2007) ) may enhance


96 cordgrass resistance to drought. Whether remnant cordgrass patches are often associated with mussels simply because mussels are common in salt marshes throughout this region (Kuenzler 1961) or because mussels are alleviating drought stress by preventing soi ls from drying out or some other mechanism (e.g. fertilizing plants to increase their tolerance to grazing) is unknown. Using surveys, model s imulations, monitoring, and field experiment s, we e xplore the role of remnant patches in mediating salt marsh tran sitions from unvegetated mudflats to cordgrass dominance and potential role of mussel mutualists in bolstering cordgrass resistance to drought. Specifically, w e test: 1. A cross the range of mudflat sizes and patch distributions that we observe in southeastern US salt marshes, do remnant patches ac celerate cordgrass recovery? 2. D oes the presence of mussels increase cordgrass survival within drought generated mudflats and, if so, is buffering against soil drying one mechanism by which mussel enhance cordgrass res istance ? 3. H ow much do remnant patches that are associated with mussel mutualists contribute to cordgrass recovery? In examining the importance of patches in hastening mudflat recovery and mussels in facilitating cordgrass during drought, this study advanc es our general understanding of patch driven recovery dynamics and mutualisms that alleviate the intensity of climate stress experienced by foundation species in controlling ecosystem resilience. Methods Study S ystem : Southeastern US Salt M arshes We condu cted the majority of this research at the National Estuarine Research Reserve on Sapelo Island , Georgia (31°24'26 "N , 81°17'24 "W ) within salt marsh platforms, spatially dominant and relatively flat expanses of marshes in this region (Schalles et al. 2013) that are monopolized by vast stands of


97 cor dgrass and flooded by the tides bet ween 0 2.5 hours twice daily. Within marsh platforms, cordgrass expands clonally via belowground ramets and emergent shoots reach ~50cm by the end of the growing season in late fall (Pennings et al 2005). At the base of cordgrass stems, mussels settle in the mud in aggregations containing between 1 to ~ 80 individuals at a density of ~ 1 aggregation every 5m 2 (Angelini et al, in preparation ).We focused our coastal survey of remnant cordgrass patches on sites located betwe en Jekyll Island, Georgia (GA) and Charleston, South Carolina (SC) as mudflats formed d uring the 2010 11 drought (Figure 5 2) in salt marshes across this geographic region. Specific coastal survey site locations were selected because they had expansive mar sh platforms and were accessible by foot. Remnant Patch Effects on Mudflat to C ordg rass T ransitions: the S alt Marsh Recovery M odel To assess the relative importance of cordgrass bordering mudflats and residing in patches in driving salt marsh recovery fr om drought, we developed a Salt Marsh Recovery (SMR) model based on Guichard et al. 2003 . In this platform in which each a cell corresponds to 0.25 m 2 of marsh platform and can be occupied by mudflat (cell value = 0), cordgrass patch (cell value = 1, each p atch occupies one cell at the start of simulations), or cordgrass border (cell value = 2, border cells frame the perimeter of mudflats at the start of simulations). To develop parameters that would allow us to simulate the recovery of real mudflats, w e use d a GPS (Trimble R6 GNSS, S unnyvale, CA) to circumscribe nine mudflats on Sapelo Island and mark the center of every cordgrass patch within each mudflat as drou ght subsided in July 2011 (Figure 5 2). To verify that


98 mudflats were generated within the timefr ame of this drought, as opposed to being relic features from prior disturbance events, we located each mudflat in Google Earth in November 2011 imagery and examined whether each unvegetated feature was present in March 2010 imagery. All nine mudflat areas were vegetated prior to the onset of this severe drought, suggesting that drought was a contributing cause of cordgrass mortality in these mudflats. We also measured the area of each patch (median patch area = 0.26m 2 ) in each mudflat and calculated the are a of each mudflat in ArcGIS. Using (Ripley 1981, Venables and Ripley 2002) , w e characterized the spatial distribution of patches within each of the six surveyed mudflats that containe d more than ten patches (see Appendix H and I for details) . From these analyses, we found that the patch distribution differed from one mudflat to another; while patches exhibited a nearly uniform distribution in one mudflat, they exhibited random and clus tered distributions in others. Based on our observations of mudflats in other salt marshes in this region (see coastal survey below), mudflats that contained no surviving cordgrass patches were also common. To estimate cordgrass expansion parameters and th us simulate mudflat recovery over realistic time intervals, we flagged 25 cm 2 permanent monitoring cells adjacent to 80 remnant cordgrass patches and 69, 25 cm 2 cells adjacent to cordgrass stands bordering mudflats in April 2011. Monitoring cells were posi tioned next to both patches and bordering stands to evaluate whether cordgrass expansion differed among the two sources of clonal growth and were distributed across the nine Sapelo Island mudflats where GPS data were


99 collected. After 12 months, we quantifi ed the density of cordgrass shoots within each monitoring cell as well as in 10, 0.25m 2 cells placed haphazardly within a scored each cell as being recovered or not (cells were rec overed if the number of cordgrass shoots in the monitoring cell shoots in healthy stand cells) and calculated the proportion of all patch and border cells that were recovered (i.e. the probability of recovering). From th ese measurements, we estimate that the probability of cells transitioning from bare mudflat to cordgrass is 0.7 if adjacent to a patch and 0.5 if adjacent to a border. Because patches expanded with similar vigor regardless of their initial size or the numb er of mussels they were associated (see Appendix J for methods, analyses, and results) and both cordgrass patches and borders advanced across mudflats with similar vigor after two and three years, we assume that: 1) all patch expand at the same rate, 2) pa tch and border expansion rate remains consistent through time, and 3) once a mudflat cell becomes colonized by a patch or border, it cannot change states. We then used the SMR model to simulate salt marsh recovery under 12 scenarios that included all comb inations of three mudflat sizes (the minimum, median, and maximum mudflat area recorded at our Sapelo field site: 23, 311, and 1,928 m 2 ), and four remnant patch distributions (uniform, random, clustered, and no patches). Although natural mudflats vary in s hape and, thus, perimeter: area ratio, we simulated cordgrass recolonization of mudflats that were square, rather than oval or some other shape, to simplify our approach. Likewise, although the initial cover of remnant patches in mudflats varied from 1.6 t o 16.1%

PAGE 100

100 cover, we set the initial cover of remnant patches to 7.3% at the start of each simulation as this was median percent cover of remnant patches recorded at in Sapelo Island mudflats. Each model simulation returns the number of time steps (years) unt 95% of mudflat cells become either a patch or border cell) as well as the final number of cordgrass border and patch cells that occupy the mudflat. We simulated mudflat recovery for each scenario three times and ana lyzed the effect size and significance of mudflat area, patch configuration, and their interaction on the recovery interval (i.e. the average number of years for mudflats to recover) using analysis of variance. The R code for the SMR model can be found in Appendix K . Regional Survey of Cordgrass Mortality and S urvival: are Mussels Enhancing Patch R esistance? To explore whether many remnant patches occur within mudflats due to the presence of mussel mutualists as our initial observations in Sapelo mudflats s uggested, we first needed to assess whether cordgrass survival was in fact higher when associated with mussels than not. To do so, we visited eleven marsh es across the Georgia and South Carolina coasts in April 2011, nine months into a severe drought (Figu re 5 2 ). At the nine sites where we observed recent and likely drought generated mudflats i.e. one or several mudflat s that contained standing dead shoots confirming recent cordgrass senescence and were characterized by dry, cracked, and/or salt crusted soil we used a transect tape to measure the mean length and width of all mudflats encountered . From these dimensions, we estimated mudflat area (A) using the equation for an oval . We did not include mudflats that abutted docks, causeways, or woody vegetat ion fringing the terrestrial marsh border or that were associated with wrack mats (i.e.

PAGE 101

101 rafts of dead plant material that covered >3% of the mudflat area) in our survey as they were unlikely to have been generated by drought (Pennings and Richards 1998) . We verified that each mudflat was vegetated prio r to the 2010 2011 drought using Google Earth imagery as described above. I n the mudflats where cordgrass survived in remnant patches , we measured the area of all cordgrass patches associated with mussels (C M ussels ) and cordgrass patches not associated wi th mussels (C No mussels ) , as well as the area of dead cordgrass associated with m ussel s (DC M ussels ) . Cordgrass patches length) mussel was observed embedded in the mud at the base of live stems that generated the patch. Based on our observation that the effect of mussels on cordgrass growth during non drought conditions is localized (i.e. 5 10cm away from a mussel or mussel aggregation cordgrass stems become less green and grow to a height similar to the background grass canopy, CA, unpublished data ), we assume the effect of mussel on cordgrass during drought is localized as well and thus requires that mussels occur within the remnant patch. Because we documented little evidence of predation (e.g. crushed shells) or heat stress (e.g. dead mussels with gaping shells) on mussels within mudflats, we expect that the distribution of mussels within mudflats did not change measurably from when cordgrass died off (~late Fall 2010) and wh en the survey data was collected (Spring 2011), meaning our DC M ussels measures are likely accurate. To summarize the relationship between remnant patches and mussels for each of the sites, we summed A, C Mussels , C No M ussels , and DC M ussels measures acro ss all mudflats surveyed and calculated the probability of cordgrass surviving when

PAGE 102

102 associated with mussels as: Mussel P Survival = (C M ussels )/ (C M ussels + DC M ussels ), and when not associated with mussels as: No Mussel P Survival = (C No M ussels )/(A C M ussels DC M ussels ). To analyze whether cordgrass survival is proportionally higher when associated with m ussel s than not, we treated Site as a random factor, M ussel Presence as a fixed factor, and analyzed the effect size and significance of M ussel Pr esence and Site on the probability of survival using a mixed effects generalized linear model using the lme4 package in R (Bates et al. 2012, R Core Development Team 2012) . Mechanism s by which M ussels E nhance C ordgrass Resistance to D rought To investigate if one mechanism by which mussels may increase cordgrass resistance to drought is that they prevent soils from drying, we ause salts become increasingly concentrated in porewater as freshwater is lost via evaporation, salinity is used commonly as a proxy for soil dryness and thus drought stress (S illiman et al. 2005) . In May 2012, we set 10 pairs of lysimeters (Angelini and Silliman 2012) to a depth of 25cm in order to span the rooting zone: one lysimeter was positioned in the middle o f a mussel aggregation (8 20 mussels per aggregation, an average size for marsh platforms) and another 1m away in a control (no mussel) marsh location. Lysimeter pairs were distributed across 6 hectares of cordgrass dominated marsh platform. Once per week from mid May through mid August, we used a refractometer to record the salinity of porewater extracted from each lysimeter. To summarize the relationship between mussels and salinity across this monitoring period, we calculated the difference in salinity between each lysimeter pair for each date, averaged these values to assess the

PAGE 103

103 mean difference in salinity across all pairs for each date, and used a t test to analyze whether the mean difference in salinity across all lysimeter pairs and throughout this time period was significantly different from zero. Because we subtracted the salinity measured in the control marsh from salinity in mussel aggregations, mean difference values that are significantly less than zero (i.e. negative) indicate salt is less con centrated in mussel aggregations than the control areas. To test if mussels, as opposed to other factors that may cause heterogeneity in soil moisture such as local changes in marsh elevation, were likely driving the salinity differences observed in our po rewater monitoring, we planted 24 cordgrass plugs, extracted from the adjacent marsh and standardized for soil volume (25 × 25 × 25cm), shoot density and height, at least 2m apart in a mudflat in April 2011 [we added mussels to plugs and transplanted plu gs into mudflat so that we could additionally test if mussels influence patch expansion (see Appendix J for methods and results)]. We haphazardly assigned each plug one of two treatments mussel addition or control (n=12 replicates per treatment) and ins erted ten mussels (length: 50 70mm) half way into the mud in a clustered pattern around the base of cordgrass stems (i.e. to mimic natural aggregations) in each addition plot. We transplanted ten mussels as this was the median density of mussels associated with cordgrass patches in our survey of GA and SC salt marsh platforms. We also agitated the top 5cm of the control plots at the start of the experiment to account for disturbance effects associated with mussel transplantation, and then left all plots und isturbed for 16 months. In August 2012, we used 0.2 × 10cm (diameter × depth) micropore rhizon samplers,

PAGE 104

104 positioned in the center of each transplant and inserted to span the top 0 10cm of the marsh, to extract porewater for salinity and ammonia measurement s. Salinity was measured using a refractometer and ammonia using standard colorimetric methods (Solorzano 1969) . We then used a t test to assess the effect size and significance of Mussel Treatment on porewater sali nity and ammonia. The Contribution of Mussel Mutualists in Driving Mudflat R ecovery To summarize the importance of the mussel mutualism for salt marsh recovery from drought, we estimated the proportion of mudflat area recovered from mussel associated pat ches (i.e. % of recovery attributed to the presence of mussel mutualists) under the 12 mudflat recovery scenarios described above (i.e. three mudflat sizes and four patch distributions, mudflat recovery was simulated under each scenario three times). To do so, we multiplied the number of patch cells that occupied the salt marsh grid the end of each simulation by 0.75 and divided this value by the number of mudflat cells that occupied the salt marsh grid at the start of the simulation. We used 0.75 as a mult iplier as this value is within the range of the proportion of cordgrass patches surviving within mudflats that were associated with mussels: 1,182 of the 1,488 (79.4%) patches observed in the GA and SC survey and 132 of the 430 (69%) patches in Sapelo Isla nd mudflats were associated with mussels. We then analyzed the effect size and significance of mudflat size, patch distribution, and their interaction on the percent of mudflat area recovered from mussel associated patches using analysis of variance.

PAGE 105

105 Resul ts Simulating cordgrass re colonization of drought generated mudflats with the SMR model revealed that mudflat size and remnant patch distribution interact to control recovery intervals (Mudflat size* Patch Distribution: F 6,24 = 359, P < 0.0001). While mudf lats are estimated to fully recover in 7 years regardless of their size if remnant patches are uniformly distributed, those with random and clustered patches take progressively longer to recover with increasing mudflat size. Clustered patches, which tend t o coalesce with one another soon after cordgrass begins to recolonize mudflats, consistently recolonized mudflats slower than either randomly or uniformly distributed patches that exhibit radial growth for longer and coalesce later in the recovery process. Across all patch distributions, the difference between mudflat with patches and those without was significant and became more pronounced as mudflats increased in size: relative to small, medium, and large mudflats with no patches that we estimate take 14, and 25 years, respectively (Figure 5 3). At the 9 South Carolina and Georgia marshes surveyed that experienced cordgrass die off, drought associated mudflats covered between 1,688 to 11,808m 2 of marsh platform area, of which 0.5 to 7.7% was covered by remnant cordgrass patches, when averaged across m udflats within sites (Appendix L ). Based on calculations derived from our survey data, the probability of cordgrass surviving the drought varied notab ly across the nine sites, but was consistently at least two orders of magnitude higher when associated with mussels [64.3 ± 27.4% (mean ± standard deviation), 28 98% (range)] than when not associated

PAGE 106

106 with mussels (1.0 ± 0.06%; 0.04 1.6%, Mussels [Sit e]: T = 6.99, P = 0.0001, Figure 5 4). Over the three months in 2012 that lysimeters were monitored to assess whether mussels may facilitate cordgrass by reducing drought stress, porewater salinity was 2 parts per thousand (ppt) lower, on average, in mussel aggregations and the adjacent control mars h (Mussels: P = 0.00019, Figure 5 5A, see Appendix M for rainfall over monitoring period ). Consistent with our salinity monitoring results, porewater salinity was 7ppt lower, on average, in the root zone of mussel addition cordgrass transplants relative to control, no mussel transplants during a relatively dry period in 2012 when salinity readings were taken (Mussel Treatment: T 1,22 = 6.9, P < 0.0001, Figure 5 5B ). In addition, mussel addition more than doubled the concentration of ammonia around the roots of experimental cordgrass transplants (Mussel Treatment: T 1,22 = 3.5, P = 0.0023, Figure 5 5C ). Finally, we found that the contribution of mussel associated patches to cordgrass re colonization depended both on th e size of the mudflat as well as the configuration of patches (Mudflat Size * Configuration: F 6,24 =12.45, P < 0.0001, Figure 5 6). In small mudflats that have a relatively high perimeter: area ratio, mussel associated patches produced 48, 41 and 35% of th e cordgrass that ultimately recolonized the mudflat if patches were distributed in uniform, random or clustered distributions, respectively, while in large mudflats the contribution mussel associated patches increased markedly and was particularly high if patches were uniformly or randomly distributed (72 and 71%, respectively, versus 64% if patches were clustered).

PAGE 107

107 Discussion Our results demonstrate that the resilience of southeastern US salt marshes to drought hinges on the presence and spatial distributi on of remnant cordgrass patches. Despite the fact that they are often quite small (<0.5m 2 ) and collectively occupy only a minor fraction of mudflats, remnant patches are able to contribute substantially to cordgrass recovery because they expand radially an d therefore colonize more area over time. Based on our monitoring and experimental results, we speculate that many cordgrass patches survived the drought due to the benefits they gained from growing in close proximity to mussels. In moderating how dry soil s become around plant roots and enhancing the availability of nitrogen, mussels are likely alleviating the primary manifestation drought associated stressors, such as snail grazing. Furthermore, despite having little effect on cordgrass expansion at the patch scale, we predict mussel mutualists dramatically enhance recovery on larger, mudflat scales through their support of associated grasses that experience enhanced siurviv orship during drought that then leads to the formation of remnant patches within a large die off area. Collectively, these findings highlight the potent, positive effect of remnant patches on the recovery of clonal growth dependent ecosystems and, more ge nerally, advance the idea that mutualists can play a keystone role in controlling climate stress and enhance the spatial distribution of sources from ecosystems can recover.

PAGE 108

108 Projecti ng S alt Marsh Recovery from Drought: are Remnant Patches I mportant? In using the SMR model to simulate cordgrass re colonization of mudflats under a range of scenarios that we observed in the field, we discovered that mudflat size and patch distribution in teract to regulate how fast salt marshes can recover and how much patches c ontribute to this process (Figure 5 3,Figure 5 6 ). This significant interaction emerges because, as cordgrass borders and patches expand inward and outward, respectively, and eventu ally coalesce, the total perimeter of borders and total perimeter of patches change accordingly to control how much cordgrass can be produce d by each source over time (Figure 5 1). In essence, mudflat size constrains the total perimeter of borders (and thu s the contribution of this source to recovery) and the spatial distribution of patches controls how quickly patches coalesce with one another (clustered patches coalesce faster than random patches which coalesce faster than uniform patches) and thus the to tal perimeter of patches as re colonization progresses (Sousa 2001) . These rather intuitive dynamics have crucial implications for the resilience of ecosystems structured by foundation species that d epend on clonal or clone like growth: while many of the mudflats that we surveyed along the GA and SC coast will likely be fully recovered in the next few years as a result of their smaller size and/or optimal distribution of patches and therefore exhibit high resilience, our model and survey studies indicate that large mudflats with no patches will likely persist for many decades and thus exhibit low resilience. Because we parameterized the SMR model with field collected data, its estimates for cordgrass r ecovery intervals and the contribution of patches to re colonization are likely fairly accurate for mudflats that will recover quickly, but

PAGE 109

109 less so for those that recover slowly. Specifically, cordgrass has been shown to advance vigorously across mudflats during years with high precipitation and nearly stall during years of moderate drought (Angelini and Silliman 2012) , suggesting that our assumption that patch and border expansion rates will re main consistent through time oversimplifies cordgrass re colonization dynamics in this system. Likewise, significant subsidence (i.e. a drop in the elevation of the marsh platform and change in soil structure due to decomposition of peat) of mudflats gener ated in a 1999 2001 drought that were still unvegetated when we observed them in 2011, indicate that cordgrass re colonization may be moderated by mudflat reinforcing feedbacks that become stronger over time ( Koppel et al. 2005) . In addition, three distinct drought s have initiated new areas of cordgrass die off in the southeastern US over the past 15 years (McKee et al. 2004, Silliman et al. 2005, Palomo et al. 2013) , hinting that drought or other disturbance events will likely inter fere with re colonization dynamics by killing off some recovered salt marsh. Consequently, we expect that the accuracy of our predications could be improved (particularly for large disturbance areas with few or no patches that will experience rainy years, subsidence, and future drought) by incorporating precipitation and time dependence into cordgrass expansion rates as well as incorporating drought disturbance into the SMR model, advancements that require further field work to define realistic parameters. Cordgrass Resistance to Drought: Extent of Mortality and the Mechanisms of S urvival In observing dry, cracked, and salt encrusted soils in recently formed mudflats six months into a severe drought, we presume that drought was the ultimate driver of cord grass mortality in the salt marsh platforms we surveyed

PAGE 110

110 along the SC and GA coasts (McKee et al. 2004, Silliman et al. 2005, Hughes et al. 2012, Palomo et al. 2013) rather than other stressors that have been implicated in previous marsh die off studies ( see Alber et al. 2008 for review ) . Across the nine sites where we observed cordgrass die off, there was notable variation in the size of d rough t generated mudflats (Appendix H ) and the distribution of remnant pat ches surviving within them (Figure 5 2). Based on our observations, we hypothesize that these differences in the extent of cordgrass mortality was driven, at least in part, by variat ion among sites in the intensity of drought (e.g. rain fell at some sites and not others, R. Phillips and D. Aren, personal communication ), elevation of marsh platforms, and the density of snails, grazers that known to interact synergistically with drought to kill cordgrass (Silliman and Newell 2003, Silliman et al. 2005) . Likewise, notable differences in the size and spatial distribution of remnant patches and proportion of those patches associated with mussels also sug gest that variation in cordgrass survival was driven, at least in part, by differences among sites in the importance of physical (e.g. heterogeneity in elevation in marsh platforms that allow some areas to be flooded by tides more than others), genetic ( e.g. stress tolerant genotypes, Hughes and Stachowicz 2004, Pandolfi et al. 2011) , and biotic (e.g. density of mussel mutualists) mechanisms that are all likely to contribute to enhancing cordg rass resistanc e to drought (Figure 5 2, Appendix L ). Across these SC and GA salt marshes that varied widely in the extent of die off, the regularity with which remnant patches were associated with mussels was striking (Figure 5 4), suggesting that conditions are commonl y less stressful, plants are generally tolerant, or both where they grow close to mussel mutualists.

PAGE 111

111 In 11 of the 14 weeks, salinity monitoring revealed that salt was more concentrated in the porewater around plant roots and thus marsh soils were outside of mussel agg regations than within them (Figure 5 5 ). Although we collected only a few measurements due to time constraints, we detected a far greater difference in salinity between mussel aggregations (range: 35 45ppt) and control soils (range: 55 80ppt) during the severe drought in 2010 than we detected in our monitoring in non drought conditions in 2012, implying that the effect of mussels can become far more pronounced drought intensifies. Results from our experimental addition of mussels provi de further evidence that mussels are providing key benefits to cordgrass that are likely leading to increased survival of adjacent stems and thus remnant patch formation. Specifically, mussel addition both increased the availability of nitrogen, and reduce d salinity below growth stunting l evels (from 45ppt to 35ppt, Figure 5 5 , Mendelssohn and Morris 2010) . The simultaneous enhancement of their growth limiting nutrient and alleviation of a major abiotic stress for cordgrass likely are leading to increased productivity of cordgrass growing immediately within mussel aggregations (Silliman and Zieman 2001, Mendelssohn and Morris 2010) and ultimately their increased resistance to death during severe drought. The mechanisms by which mussels increase nutrients lik ely include their deposition of pseudofeces and facilitation of high densities of infaunal invertebrate that excrete nitrogen rich waste (Bertness 1985, Bertness and Grosholz 1985) . Comparatively, the mechanism by which mussels maintain hi gher soil moisture that the adjacent compact, salt marsh platform is not as intuitive, but likely includes their paving of the marsh surface with their hard

PAGE 112

112 shells and deposition of a thick layer of pseudofeces that may slow evaporative water loss and faci litation of crabs that may create water storage compartments belowground through their excavation of burrows. Given that drought elicits a cascade of other biogeochemical and ecological changes, including low pH and high metal concentrations in soils (McKee et al. 2004, Palomo et al. 2013) and enhanced impacts of snail grazing (Silliman et al. 2005) and mussels (and the organisms they facilitate) likely counteract these changes in multiple ways, the mechanisms driving the pattern of enhanced cordgrass survival on mussel aggregations are likely more complex than we have examined in this study. Additional research to expose the relative importance of different mechanisms by which mussels facilitate cordgrass will not only deepen our understanding of south eastern US salt marshes, but also likely provide insight to the resilience of the many other coastal ecosystems where bivalves overlap with vascular plant foundation species (e.g. mangroves, intertidal and subtidal seagrasses). Effect of Mussel Mutualists on Cordgrass Recovery at Patch and Mudflat S cales Our experimental addition of mussels and monitoring of natural patch expansion revealed that neither the presence nor density of mussels have a detectable effect on cordgrass expansion at the patch scale, despite their localized effects on moderating soil stress and increasing nutrient availability (Appendix J ). Given that ample rain fell on Sapelo Island m arshes as drought subsided (Figure 5 2 ) and decomposition of dead plant material likely created pools of nutrients in mudflat soils, it makes sense that mussels had a negligible effect on cordgrass re colonization at the patch scale. When considered at a mudflat scale, however, the potential contribution of mussel mutualists is

PAGE 113

113 impressive: we estimate that between 35 73% of the cordgrass that ultimately recolonizes mudflats is generated by mussel associated patches (Figure 5 6). Given that remnant patches contribute progressively more to recovery as mudflats increase in size and as patches are more uniforml y spread, this mutualism is likely enhancing salt marsh resilience the most where disturbances are particularly large and where physical and environmental factors, such as food availability and the supply of mussel larvae, cause mussel aggregations to be o ptimally distributed across salt marsh platforms. In ecosystems that may be vulnerable to large scale disturbances and do not have significant cover of stress alleviating mutualists due to low recruitment, our results suggest that resilience may be proacti vely enhanced by simple management strategies that enhance the spatial distribution of mutualists across stands of habitat forming foundation species. Conclusions Collectively, our results integrate two concepts that have long been emphasized in studies of plant and community succession and more recently been gaining attention in studies on ecosystem resilience: positive interactions and remnant patch dynamics (Levin and Paine 1974, Yarranton and Morrison 1974, Paine and Levin 1981, Sousa 1984, Halpern et al. 2007, Angelini and Silliman 2012, Zahawi and Holl 2013) . Given that positive interactions occur in diverse ecosystems (Bertness and Callaway 1994, Bruno et al. 2003, Brooker et al. 2008) and foundation species are more likely to recover quickly, regardless o f their reproductive mode, if many mature individuals survive within a disturbed area, we antic ipate that mutualisms that sustain patches of adult individuals and

PAGE 114

114 thus give rise to remnant patch recovery dynamics play a keystone role in controlling the res ilience of many other ecosystems. E vidence of mutualism breakdown as a result of climate change and human modification of food webs and nutrient dynamics warn that these positive interactions have thresholds beyond which they fail (Kiers et al. 2010) and are therefore unlikely to provide blanketing security against disturbance, especially as climate continues to change and humans continue to overuse natural resources and degrade ecosystem structure . Consequently, actions to comba t these ultimate drivers of ecosystem collapse climate change, biodiversity loss, eutrophication, and other human driven factors are needed even more than proactive management actions such as enhancing mutualisms to effectively enhance ecosystem resilien ce.

PAGE 115

115 Figure 5 1. Remnant patch dynamics in southeastern US salt marshes . A) The pattern and pace with foundation species can re colonize disturbed areas (in white) from bordering stands (light green) and remnant patches (dark green) can vary depending on the spatial distribution of those patches. B) In the southeastern US, drought caused mass mortality of cordgrass in salt marshes, cr eating largely bare mudflats that were interspersed with C) remnant cordgrass patches , many of which are associated with D) aggregations of the ribbed mussel, Geukensia demissa .

PAGE 116

116 Figure 5 2. Drought trends over study period . Monthly summary of drought intensity, measured by the Palmer Drought Severity Index (PDSI) from January 2007 through December 2013 for the South Car olina coastal region (in dark grey) and Georgia coastal region (in black) . The periods over which field experiments and monitoring were conducted are indicated by brackets and abbreviated titles see methods for further details . PDSI values that fall below 0 indicate relatively dry years for this region according to the historical climate record, and values that fall below 2 and 4 indicate moderate and severe drought conditions, respectively (source: NOAA National Climatic Data Center, htt p://www.ncdc.noa web/) . Likewise, values that fall above 0 indicate relatively wet years.

PAGE 117

117 Figure 5 3. Salt Marsh Recovery model results. The effect of mudflat size (on x axis), and remnant patch distribution (no patches, uniform, random, clustered) on the nu mber of years for cordgrass to recolonize mudflats. Points denote the average recovery interval and contribution of mussel associated cordgrass patches to recovery of three simulation runs. Standard error bars are too small to be seen given the scale of th is figure.

PAGE 118

118 Figure 5 4. Coastal survey of remnant patch association with mussels. Location of nine marshes where extensive drought generated marsh die off was observed and the probability cordgrass surviving within these mudflats when associated with m ussels (black bars) or growing alone (grey bars). Note the break and scale in the x axis.

PAGE 119

119 Figure 5 5. Relationship between mussels and salinity. A) Summary of differences in porewater salinities in mussel aggregations than in the adjacent control (n o mussel) marsh over 14 weeks of monitoring in 2012 and effect of experimental addition of mussels on B) porewater salinity and C) ammonia concentratio n in the cordgrass root zone . Salinity values in corresponds to the average difference for each weekly mo nitoring date for 10 lysimeters paired on and off of mussel aggregations. Data are shown as the mean ± SE of 12 replicat e transplants per treatment in B and C .

PAGE 120

120 Figure 5 6. Estimated contribution of mussel associated remnant patches to mudflat recovery . Effect of mudflat size (on x axis) and patch contribution (in different line types and symbols) on the estimated contribution of remnant patches associated with mussels on mudflat recovery. Data points indicate the mean of three model simulations and sta ndard errors are too small to be visible on this figure.

PAGE 121

121 APPENDIX A MECHANISMS OF CORDGRASS FACILITATION OF MUSSELS To assess whether cordgrass alleviates temperature stress to facilitate mussels in southeastern US salt marshes as shown in New England f ringing salt marshes (Altieri et al. 2007) and identify whether interactions among cordgrass, mussels and invertebrates ar e an example of a facilitation cascade, we conducted a field experiment in a Sapelo Island marsh platform in July 2011. Specifically, we marked 24 plots spaced >1m apart in a recently formed mudflat and planted a cordgrass transplant of a standard size (20 cm 3 plug of roots + marsh soil and 8 10 cordgrass stems) in each. The experiment was conducted in a mudflat, rather than in a healthy stand of cordgrass, to prevent neighboring vegetation from influencing the efficacy of experimental treatments. Each plot was then haphazardly assigned one of four treatments: Canopy + Shade, Canopy + No Shade, No Canopy + Shade, or No Canopy + No Shade. The live cordgrass stems (i.e. the canopy) was left unmanipulated in Canopy plots and removed with shears in No Canopy plot s. Experimental shades constructed using PVC corner posts and a sheet of landscape fabric were placed 40cm above Shade plots, while No Shade plots were left unmanipulated. PAR and iButton temperature readings collected at the marsh surface in Shade plots a nd healthy cordgrass monocultures in the marsh platform indicate that Shades effectively mimicked the shading provided the cordgrass canopy at this elevation (i.e. 70% light reduction). We then transplanted 10 mussels (size: 50 70mm shell length) into each plot in a cluster to mimic natural aggregations. After 1, 3, 7 and 10 days, we scored mussel survival. Mussel mortality occurred between days 7 and 10 during a summer heat wave that was within typically of the southeastern US.

PAGE 122

122 Figure A 1. Mechanisms of cordgrass facilitation of mussels. A) Effect of cordgrass canopy and shade presence on the survival of ribbed mussels over 10 days in July 2011. Treatments are denoted by different colors and symbols and data are shown as the mean ± SE of 6 replicate plots per exceeded 43 ° C on the mud surface is noted by the red box. Also shown is B) an example of the e xperimental shades and D) mussels that have died during the heat shock.

PAGE 123

123 APPENDIX B RELATIONSHIP BETWEEN MUSSEL AGGREGATION AREA AND MUSSEL DENSITY We used Analysis of Variance (ANOVA) to examine the effect size and significance of Site on the average aggregation area and Analysis of Covariance (ANCOVA) to assess whether the slope of the relationship between aggreg ation area and the number of mussel in the aggregation differed among sites (n=17 aggregations measured per site) . Although the average aggregation area (mean [SD]: 0.039 [0.080] and 0.038 [0.67]) is similar at the two sites (Site: P=0.38, Fig. 2), the rel ationship between aggregation area and the number of mussels is best fit by a power law function which differs significantly between field sites (ANCOVA: Site: P = 0.04). Fig ure B 1. Relationship between the area of a mussel aggregation area and the nu mber of mussels it contains .

PAGE 124

124 APPENDIX C SOIL DEPOSITION MEASUREMENT To verify whether our method for quantifying soil accretion (i.e. inserting a small PVC rod into the marsh until it contacted the root mat) reflects differences in soil deposition, we s ecured a 7 cm diameter filter paper to the marsh with PVC anchor pins in a position that would not obstruct the filter feeding activity of mussels or damage cordgras s stems within each plot (Smith and Frey 1985, Hensel et al. 1998) . Afte r 48 hours, we collected, oven dried, and weighed the papers. We found that our shor t term soil deposition measures was positively and significantly correlated with soil accretion depth (R 2 =0.46, P=0.0002), indicating our measures provide consistent eviden ce that mussels cause more sediment to accumulate on the marsh as aggregations become more dense . Figure C 1 . Summary of the effect of mussel density on soil deposition rate .

PAGE 125

125 APPENDIX D SOIL AMMONIA AND ORGANIC CARBON MEASUREMENTS To evaluate whether the short term bait lamina test provided a reliable measure of decomposition, we collected four replicate (4.5 × 10cm, diam. × depth) soil cores and extracted porewater using rhizons from the top 10 cm of the marsh to measure organic Carbon and ammonia co ncentrations, respectively. Due to time constraints, only 17 of our 24 plots were measured for porewater ammonia. A standard loss on ignition method was used to measure the p ercent organic Car bon in the soil (Craft et al. 1991) and the phenolhypochlorite method was used to measure ammonia in the porewater (Solorzano 1969) . Because organic Carbon is consumed an d ammonia produced during decomposition , their relative concentrations in the soil should decrease and increase, respectively, with increasing levels of decomposition. T he decomposition of bait holes negatively correlated with percent soil organic Carbon ( R 2 = 0.20, P= 0.02) and positively correlated with porewater ammonia (R 2 = 0.21, P=0.05 , Fig A3), providing consistent lines of evidence that decomposition increases with increasing mussel density.

PAGE 126

126 Figure D 1 . Relationship between mussel density and two measures related to decomposition. A) Effect of musse l density on soil organic C and B) porewater NH 4 + .

PAGE 127

127 APPENDIX E SUMMARY OF NULL, LINEAR, LOG, HYPERBOLIC, AND POWER MODEL COMPARISONS Table E 1. Summary of five models describing the relationship betwee n M , the number of mussels added, and Y , each respo nse variable. The best fit models for each relationship are those with the lowest AICc value . Response Parameters estimates Model Types A B Z AICc AICc wt Functional group richness Null, y= a 4.67 NA NA 72.0 30.04 1.91 * 10 7 Linear, y=a + bM 4.16 0.026 NA 60.2 18.27 6.87*10 5 Log, y=a +b*log(M+1) 3.35 0.61 NA 42.0 0 0.64 Hyperbolic, y= a*M/(b + M) 5.43 0.52 NA 83.9 41.92 5.03*10 10 P ower, y=a+ bM z 2.66 1.09 0.23 43.1 1.13 0.36 D Null, y= a 1.7 NA NA 36.8 8.78 0.005 Linear, y=a + bM 1.48 0.011 NA 28.0 0 0.43 Log, y=a +b*log(M+1) 1.25 0.21 NA 28.1 0.07 0.41 Hyperbolic, y= a*M/(b + M) 2.01 0.82 NA 54.3 26.26 8.49*10 7 P ower, y=a+ bM z 1.36 0.075 0.58 30.1 2.04 0.15 Mud crabs Null, y= a 0.67 NA NA 61.9 15.34 1.7*10 4 Linear, y=a + bM 0.23 0.022 NA 47.9 1.28 0.19 *Log, y=b*log(M+1) 0 0.33 NA 46.6 0 0.36 Hyperbolic, y= a*M/(b + M) 2.56 37.29 NA 48.5 1.91 0.14 *Pow er, y= bM z 0 0.16 0.56 46.9 0.29 0.31 Marsh crabs Null, y= a 0.75 NA NA 93.2 8.75 5.0 *10 3 Linear, y=a + bM 0.032 0.036 NA 84.7 0.21 0.36 *Log, y=b*log(M+1) 0 0.42 NA 85.6 1.08 0.23 Hyperbolic, y= a*M/(b + M) 0.37 7.28 NA 97.4 12.88 6.4*10 4 * Power, y=a+ bM z 0 0.074 0.83 84.5 0 0.40 Snails** Null, y= a 183.92 NA NA 259.7 0.81 0.34 Linear, y=a + bM 197.7 0.69 NA 258.9 0 0.52 Log, y=a +b*log(M+1) 197.78 6.37 NA 261.5 2.63 0.14 Hyperbolic, y= a*M/(b + M) 6.21 41.21 NA 323.5 64.61 4.8* 10 15 Adult fiddler crabs Null, y= a 7.08 NA NA 116.3 2.25 0.16 Linear, y=a + bM 7.77 0.035 NA 115.6 1.54 0.23 Log, y=a +b*log(M+1) 8.73 0.76 NA 114.1 0 0.51 Hyperbolic, y= a+ M/ (b + M) 6.14 0.38 NA 136.2 22.13 7.9*10 6 Power, y=a+ bM z 8.5 4 0.55 0.41 117.4 3.33 9.5*10 3 Juvenile fiddler crabs* Null, y= a 28.08 NA NA 241.5 33.45 2.9*10 8 Linear, y=a + bM 5.29 1.15 NA 208. 1 0 0.53 Log, y=a +b*log(M+1) 0 14.47 NA 222.7 14.66 3.4 *10 4 Hyperbolic, y= a*M/(b + M) 373.3 232.8 NA 208 . 9 0.81 0.35 Power, y=a+ bM z 4.89 1.28 0.97 210. 8 2.9 0.12 * y intercept set to 0 ; ** Power function was not fit ; *** Hyperbolic function was not fit

PAGE 128

128 Table E 2 . Summary of five models describing the relationship betw een the number of mussels added (M) and each ecosystem function response ( Y ) . Response Variable Parameter estimates Model A B Z AICc AIC AICcwts Soil Accretion Null, y= a 3.3 NA NA 104.9 44.15 1.3*10 10 Linear, y=a + bM 2 0.07 NA 73.9 13.13 7.3*10 4 Log, y=a +b*log(M+1) 0.49 1.29 NA 60.8 0 0.52 Hyperbolic, y= a*M/(b + M) 7.08 10.42 NA 64.8 4.08 6.8*10 3 Power, y=a+ bM z 0.89 0.77 0.46 61.2 0.48 0.41 Infiltration** Null, y= a 15.45 NA NA 231.2 20.03 3.8*10 5 Linear, y=a + bM 0 0.78 NA 211.1 0 0.85 Log, y=a +b*log(M+1) 0 8.56 NA 221.1 9.93 5.9*10 3 Power, y=a+ bM z 6.34 0 . 00007 3 .14 214.6 3.45 0.15 Decomposition N ull, y= a 0.61 NA NA 20.3 5.02 3.8*10 2 Linear, y=a + bM 0.55 0.0029 NA 25.3 0 0.46 *Log, y=b*log(M+1) 0.50 0.052 NA 24.9 0.47 0.37 Hyperbolic, y= a*M/(b + M) 0.66 0.18 NA 1.1 26.41 8.5*10 7 *Power, y= bM z 0.50 0.044 0.4 22.8 2.49 0.13 Aboveground cordgrass biomass Null, y= a 68.48 NA NA 220.8 6.51 2.4*10 2 Linear, y=a + bM 57.64 0.53 NA 216.2 1.97 0.23 *Log, y=b*log(M+1) 44.81 10.7 NA 214.3 0 0.63 Hyperbolic, y= a*M/(b + M) 17.08 8.13 NA 263.2 48.96 1.5*10 11 *Power, y=a+ bM z 49.08 5.82 0.48 217.7 3.44 0.11 Belowground cordgrass biomass Null, y= a 15.48 NA NA 92.2 0 0.59 Linear, y=a + bM 15.29 0.0094 NA 94.4 2.12 0.20 Log, y=a +b*log(M+1) 15.39 0.042 NA 94.9 2.63 0.16 Hyperbolic, y= a* M/(b + M) 15.66 0.12 NA 157.4 65.11 4.3*10 15 Power, y=a+ bM z 15.31 0.0027 1.29 97.3 5.05 4.7*10 2 Benthic algae biomass Null, y= a 16.56 NA NA 104.9 0 0.39 Linear, y=a + bM 16.33 0.01 NA 107.1 2.24 0.13 Log, y=a +b*log(M+1) 15.47 0.49 NA 104.9 0.06 0.38 H yperbolic, y= a+ M/ (b + M) 17.51 0.19 NA 152.6 47.71 1.7*10 7 Power, y=a+ bM z 14.83 1.43 0.13 107.5 2.61 0.11 Invertebrate biomass Null, y= a 28.08 NA NA 241.53 33.45 2.9*10 8 Linear, y=a + bM 5.29 1.15 NA 208.08 0 0.53 Log, y=a +b*log(M+1) 0 1 4.47 NA 222.74 14.66 3.4 *10 4 Hyperbolic, y= a*M/(b + M) 373.3 232.8 NA 208.89 0.81 0.35 Power, y=a+ bM z 4.89 1.28 0.97 210.98 2.9 0.12 * y intercept set to 0 ; ** Power function was not fit ; *** Hyperbolic function was not fit

PAGE 129

129 T able E 3 . Summary of five models describing the effect mussels on multifunctionality, where M refers to the number of mussels added, Y refers to each response variable. Parameters estimates Response Variable Model a B z AICc delta AIC AICc weights Average Multifu nctionality Null, y= a 53.77 NA NA 187.0 45.96 8.5*10 11 Linear, y=a + bM 44.46 0.45 NA 144.0 2.93 0.19 Log, y=a +b*log(M+1) 36.04 8.02 NA 152.5 11.43 2.7*10 3 Hyperbolic, y= a*M/(b + M) 65.27 1.03 NA 203.8 62.78 1.9*10 14 Power, y=a+ bM z 40.91 2 .26 0.64 141.1 0 0.81 Threshold Indices: the number of functions performed above a % of maximum functioning >10% threshold Null, y= a 6.30 NA NA 34.2 15.7 2.4*10 4 Linear, y=a + bM 6.04 0.013 NA 18.5 0 0.62 Log, y=a +b*log(M+1) 5.80 0.23 N A 21.3 2.8 0.15 Hyperbolic, y= a*M/(b + M) 6.52 0.12 NA 108.9 90.4 1.5*10 20 Power, y=a+ bM z 5.95 0.054 0.68 20.6 2.1 0.22 >30% threshold Null, y= a 4.96 NA NA 87.7 19.8 2.7*10 5 Linear, y=a + bM 4.16 0.04 NA 70.0 2.1 0.19 Log, y=b*log(M+1) 3.30 0.76 NA 67.9 0 0.54 Hyperbolic, y= a*M/(b + M) 6.25 1.37 NA 101.4 33.5 2.9*10 8 *Power, y= bM z 3.66 0.33 0.54 69.3 1.4 0.27 >50% threshold Null, y= a 4.04 NA NA 91.9 27.9 4.2*10 7 Linear, y=a + bM 3.08 0.049 NA 63.8 0 0.49 Log, y=b*log(M+1) 2.1 8 0.86 NA 68.8 5.1 3.9*10 2 Hyperbolic, y= a*M/(b + M) 6.25 4.37 NA 93.9 30.1 1.4*10 7 *Power, y=a+ bM z 2.74 0.21 0.67 63.8 0.04 0.48 >70% threshold Null, y= a 2.96 NA NA 93.4 22.7 8.7*10 6 Linear, y=a + bM 2 0.048 NA 70.7 0 0.74 Log, y=a +b*log (M+1) 1.17 0.82 NA 76.2 5.5 4.7*10 2 Hyperbolic, y= a* M/(b + M) 5.09 5.97 NA 86.1 15.4 3.3*10 4 Power, y=a+ bM z 1.81 0.13 0.78 73.2 2.5 0.21 >90% threshold Null, y= a 1.04 NA NA 85.51 27.62 8.9*10 7 Linear, y=a + bM 0.23 0.041 NA 61.97 4.08 0.12 Log, y=a +b*log(M+1) 0 0.5 NA 75.36 17.47 1.4*10 4 Hyperbolic, y= a+ M/ (b + M) 0.24 4.12 NA 96.81 38.92 3.1*10 9 Power, y=a+ bM z 0.52 0.000037 2.62 57.9 0 0.88

PAGE 130

130 Table E 4 . Summary of best fit model equations predicting the relationship between t he number of mussels in an aggregation (M) and number of functions exceeding a range of percent of maximum functioning threshold values (Y). Mussel treatment significance Best fit model T slope T exponent P slope P exponent Threshold Index % 10% Y = 0.013M + 6.04 5.06 NA <0.0001 NA 30% Y = 0.76*log(M+1) + 3.30 5.84 NA <0.0001 NA 50% Y = 0.049M + 3.08 7.52 NA <0.0001 NA 70% Y = 0.048M + 2.00 6.42 NA <0.0001 NA 90% Y = 0.052M 2.62 + 0.00004 0.21 2.35 0.8397 0.0286 T able E 5 . Summary of best f it model equations predicting the relationship between the effective number of functional groups , D , and both average multifunctionality and the number of functions exceeding a range of percent of maximum functioning threshold values (Y). Functional group diversity= D (i.e. e ) Average multifunctionality Y = 0.29e + 2.72 7.34 NA <0.0001 NA 10% threshold Y= 0.92e + 4.71 4.97 NA <0.0001 NA 30% threshold Y= 0.65(log e +1)+ 2.65 7.97 NA <0.0001 NA 50% threshold Y= 1.19*log(e +1)+ 5.61 7.39 NA <0.0001 NA 70% threshold Y= 5.52*log(e +1)+ 0.11 5.29 NA <0.0001 NA 90% threshold Y= 8.73*10 7 (e ) 1.73 + 0.61 0.15 2.31 0.88 0.03 Figure E 1 . Effect of invertebrate functional group diversity, measured as e (Diversity) , on the number of functions excee ding a range of threshold values, from 10 to 90% of maximum functioning. Points correspond to individual experimental plots and line fits reflect the best fit model according to model comparison tests (see Table E 5 for line equations).

PAGE 131

131 APPENDIX F FREQUEN CY OF MUSSEL AGGREGATIONS OF DIFFERENT SIZES AT TWO SAPELO ISLAND SALT MARSH PLATFORMS Figure F 1. Mussel distribution survey. The aggregate area frequency distribution of mussel aggregations observed in 500 m 2 areas surveyed in Sapelo Island, GA marsh platfor ms .

PAGE 132

132 APPENDIX G CORDGRASS RECOVERY FROM SEEDS AND CLONAL GROWTH: AN EXPERIMENT To deter mine whether seed dispersal , clonal expansion, or both are reproductive modes by which cordgrass re colonizes mudflats , we marked two, 1 m 2 plots in a mudflat i mmediately adjacent to each of ten surviving cordgrass patches in May 2008. An intensive drought lasting from 2 007 to early 2008 generated the ~ 1,200m 2 mudflat within which this experiment took place (CA and J.vK, personal observation ) . We assigned one pl ot in each pair one of two treatments: clonal ramet exclusion or control. To prevent cordgrass from colonizing via clonal growth and thus isolate the contribution of seeds to recovery, we used a flat shovel to install landscape fabric around the perimeter of each runner exclusion plot to a depth of 30cm, a depth below which cordgrass produces very few rhizomes (LPML, AJPS, TvH, MH, CA, unpublished data ). Control plots were establish ed 1m from each ramet exclusion and trenched with the shovel to account for disturbance effects . In July 2009, we determined percent cordgrass cover using a 100 cell frame positioned over each plot. In July 2010 and October 2011, we modified our method to better assay the source of emergent shoots: in each unit, we tugged on each shoot to test whether it was anchored by a deep penetrating rhizome, indicating it was a ramet, or not, indicating a seedling, and recorded the number of each. S hoots that emerged in the ramet exclusion identified as ramets were removed. We assigned the Pa tch as a random, blocking factor and Ramet Treatment as a fixed factor and analyzed the effect size and significance of each and their interaction using a mixed effects ANOVA on the percent cover data from 2009. As we did not

PAGE 133

133 observe a single cordgrass see dling in any plot over the duration of the experiment, multivariate analyses that incorporate responses of both cordgrass seeds and ramets were unnecessary. Consequently, we analyzed the effect size and significant of Patch (a random effect), Ramet Treatme nt, and their interaction over time on the number of ramets per plot using repeated measures ANOVA. After one year, we found that control plots enervated by cord grass ramets had recovered to 33 ± 4 (mean ± SE) % cover while those wi th runners remained bare (Clonal Ramet Treatm ent t test: P <0.0001 ). Similarly, after 2 and 3 years, we counted an average of >100 live shoots in ramet control plots while runner exclusions had zero seedlings (Clonal Ramet Treatm ent * Time repeated measures ANOVA; Time:T= 4.09, P =0.0003 Treatment * Time: T=2.67, P <0.0001, Figure G 1 ) .

PAGE 134

134 Figure G 1 . Effect of c ordgrass clonal ramets and seeds (grey) and seeds only (i.e. ramet exclusion in black and zeros) on A) the percent cover of c ordgrass after 1 year and B) on the number of cordgrass live shoots after 2 and C) 3 years in plots positioned in die off mudflats adjacent to surviving cordgrass patches. Data are shown as the mean ± SE of 10 replicate plots per treatment.

PAGE 135

135 APPENDIX H SAPELO ISLAND MUDFLAT CHARACTERISTICS Table H 1 . Characteristics of 9 Sapelo Island, GA mudflats. Summary of GPS data collected in 2011 on the size of recently formed mudflats and the distribution and cover of cordgrass only patches and cordgrass patches associated with mussels. These data were used to parameterize the SMR model. ID # Mudflat (m 2 ) Border perimeter (m) # cordgrass only patches # cordgrass patches with mussels % Initial Patch Cover 1 22.888 27.663 0 1 4.61 2 37.21 25.246 1 5 2.62 3 59.27 32.577 4 7 6.92 4 134.81 50.317 0 22 10.04 5 311.436 80.31 18 55 3.73 6 372.717 102.98 43 4 10.61 7 380.22 95.209 31 30 16.30 8 534.416 143.388 4 21 7.31 9 1928.96 387.55 31 153 15.42

PAGE 136

136 APPENDIX I SPATIAL DISTRIBUTION OF PATCHES IN SAPELO ISLAND MUDFLATS Figure I 1. Summary of the remnant patc h distribution in 4 mudflats. Examples of the spatial distribution of cordgrass patches surviving within mudflats confidence intervals, indicated by the dotted lines, indicate patches are clustered at the spatial scale indicated on the x axis (units in m) and values that fall below this interval indicate patches are over dispersed at that spatial scale, while values in the middle indicate patches are randomly distributed.

PAGE 137

137 APPENDIX J EFFECT OF MUSSELS ON CORDGRASS RECOVERY AT THE PATCH SCALE Methods To assess if mussels influenced the vigor with which cordgrass expands across mudflats and therefore influences re c olonization rates at the patch scale, we counted the number of cordgrass shoots emerging within a 25 × 25cm frame positioned along the initial border of each experimental mussel addition transplant (i.e. expansion shoot density) and measured the distance b etween the center of each transplant and the furthest cordgrass shoot observed (i.e. radial growth) and repeated these measurements in four locations around each transplant see Methods in main text for full experimental design. To contextualize the result s of our transplant experiment and explore whether the number of mussels associated with cordgrass patches regulates their radial growth, we marked the center and numbered 80 patches located within recently formed, drought generated mudflats on Sapelo Isla nd. For each patch, we counted the number of visible mussels, if any, associated with the patch and measured its radius in April 2010, and again in both April 2011 and April 2012 using the method described above. We did not collect data on patches that coa lesced with cordgrass stands that bordered mudflats or other patches as we could not determine the source of emergent shoots: consequently, we have data on 80, 71, and 64 patches in years 2010, 2011, and 2012, respectively. We calculated the lateral expans ion, Y , of each patch, i, as the difference in the average patch radius, R i , over the first year (Y i = R i , 2011 R i , 2010 ), second year (Y i = R i , 2012 R i , 2011 ), and overall (Y i = R i , 2012 R i , 2010 ). Since patch size can mediate

PAGE 138

138 cordgrass expansion (Angelini and Silliman 2012) and may therefore interact with mussel density to control patch radial growth, we used multiple linear regression to investigate the individual and interactive effects of mussel density and initial patch area on radial growth in R (R Core Development Team 2012) . Results Although mussels significantly reduced porewater salinity and increased ammonia, these seemingly bene ficial changers were not associated with a concomitant increase in cordgrass transplant expansion. Instead, transplants expanded a similar distance across mudflats and produced a similar number of shoots in the expansion zone regardless of mussel presence (Mussel Treatment: T 1,22 1 ). In following, monitoring of naturally occurring patches revealed that neither the number of mussels that associate with patch nor the initial size of a remnant patch had a significant effect on the la teral expansion of cordgrass patches after 1 or 2 years or on the expansion between 0.12). Over one year, experimental cordgrass transplants expanded 27 ± 2cm (mean ± SE, range : 8.5 to 46cm), while naturally occurring patches expanded 29 ± 3 cm (range: 1 to 127cm), on average.

PAGE 139

139 Figure J 1. Effect of mussels on patch expansion. Effect of experimental addition of mussels on A) the growth of cordgrass patches transplanted into m udflats after 16 months, measured in terms of radial expansion and B) the number o f cordgrass expansion shoots and C) growth of natural cordgrass patches t hat survived within mudflats after 1 (dark grey points) and 2 years (light grey points) as well as D) a summary of path analysis conducted on na tural patch expansion results . Data are shown as mean ± SE of 12 replicat e transplants per treatment in A D .

PAGE 140

140 Implications Although experimental addition of mussels increased ammonia concentrations, a nutrient tha t typically limits cordgrass growth in southeastern US salt marshes (Silliman and Zieman 2001) , it did not have a detectable effect on cordgrass patch expansion. Likewise, neither the number of mussels a patch was associated with nor its initial size drove predictable variation in natural patch expansion over two ye ars. Together, these experimental and correlational results indicate that remnant patch expansion is not likely limited by nutrients or water, i.e. resources that mussels modify, over the period of time that we monitored growth. Given that ample rain fell on Sapelo Island mar shes as drought subsided and decomposition of dead plant material likely created pools of nutrients in mudflat soils, it makes sense that mussels had a negligible effect on cordgrass re colonization at the patch scale. Importantly, 78 o f the 80 natural patches we monitored recolonized progressively more mudflat over time through their radial expansion, so much so that many (20%) had coalesced with another patch or border after only two years.

PAGE 141

141 APPENDIX K R CODE FOR THE SPARTINA RECOV ERY MODEL # Salt Marsh Recovery Model # Christine Angelini & Johan van de Koppel April 2014 # First setup of the model remove(list=ls()) # Remove all variables from memory on=1;off=0; # Declaration of the terms on and off require("simecol") # Loading package simecol http://simecol.r forge.r require("R.matlab") # A package that allows the data to be saved to matlab, for movie making setwd("C:/Users/ ") # Set this to the folder containing the scripts below # Loading the func tions that create the fractal and regular patches functions source('RandomFractalPatches.r', echo=FALSE) source('RegularPatches.r', echo=FALSE) # ----Settings of the simulation # Remaining vegetation patches distribution, being "Random", "Fractal", "Even ", "Bare" ## See additional code to get Fractal or Even patches below Patches = "Fractal" Border = on; Bare = 1 Patch = 2 Borders = 3 # ----Model parameters ----------------------------------------------------Alpha1 = 0.5 # Probability of bare cell becoming colonized by a border, if all neighbors are colonized Alpha2 = 0.7 # Probability of bare cell becoming colonized by a patch, if all neighbors are colonized FractionPatch=0.073 # Fraction of cells occupied by patches CoverThreshol d = 1 # If the cover goes beyond this value, the return time is measured # Simulation parameters n = 12 # Dimensions of the simulated landscape EndTime = 60 # Number of timesteps in the whole simulation NoFrames= EndTime # Number of frames displayed during the entire simulation # Initialisation of the variables wdist8=matrix(nrow=3,c(1,1,1,1,0,1,1,1,1)) Neighbourhood=wdist8 # Matrix of influence for growth # Matrix of influence of borders

PAGE 142

142 NumberNeig hbors=sum(Neighbourhood) # Initial values of the Cell matrix # Building the matrix containing the vegetation is declaired (3 = patch, 2 = veg, 1 = bare) Cells = matrix(ncol=n,nrow=n, data=1) # An empty stretch of marsh if (Patches=="Random"){ # random nu mber for each cell PatchVector = (runif((n 2)*(n 2))
PAGE 143

143 for (Time in 1:EndTime){ SumBorderOccu=matrix(nrow=n,neighbours(x=Cells,state=3,wdist=Neighbourhood)) SumPatchOccu=matrix(nrow=n,neighbors(x=Cells, state=2,wdist=Neighbourhood)) NBord=SumBorderOccu/NumberNeighbors # NBord is the proportion of neighbors that are Borders NPatch=SumPatchOccu/NumberNeighbors #NPatch is the proportion of neighborsPatch Rb=matrix(nrow=n,data=runif(n*n)) # random number for each cell Rp=matrix(nrow=n,data=runif(n*n)) # random number for each cell ColonisationBorder= 2*((Cells==Bare)&(Rb<=(NBord*Alpha1))&((NBord*Alpha1)>(NPatch*Alpha2))) ColonisationPatch= (Cells==Bare)&(Rp<=(NPatch*Alpha2))&((NPatch*Alpha2)>(NBord*Alpha1)) # Combining it all the calculate the new cell state Cells = Cells + ColonisationBorder + ColonisationPatch # Graphic representation of the model every now and then if (ii>=EndTime/NoFrames){ image(Cells, zlim=c(1,3), xaxt="n", yaxt="n", col = c("black", "dark green","green"), sub = "Dark Green = patch;Green=Patc h; Black=bare", add = TRUE) #quartz.options(title=paste("Time : ",sprintf("%20.0f",Time), # "of" ,sprintf("%20.0f",EndTime), "years")) Cells_in_Time[,,jj] = Cells; ii=0 # Resetting the plot counter jj=jj+1 # Increasing the graph update counter by 1 } ii=ii+1 # The plot counter is updated PatchCells_in_Time[Time]=sum(Cells==2) #to measure patch growth over time BareCells_in_Time[Time]=sum(Cells==1) #to measure the initial size of bare mudflat and rate o f closure over time BorderCells_in_Time[Time]=sum(Cells==3) #to measure border advancement PatchCover_in_Time[Time]=sum(Cells==2)/n/n BorderCover_in_Time[Time]=sum(Cells==3)/n/n TotalCover_in_Time[Time]=(PatchCover_in_Time[Time]+BorderCover_in_Ti me[Time]) if(TotalCover_in_Time[Time]>=CoverThreshold && ReturnTimeTrigger==on) {ReturnTime=Time ReturnTimeTrigger=off;}} plot(1:EndTime,TotalCover_in_Time, xlab="Time (yr)") if(is.nan(ReturnTime)) { cat('The vegetation did not recover')} else { cat(sprintf('Return time: %1.0f years', ReturnTime))}

PAGE 144

144 write.csv(t(BareCells_in_Time), file='XXX.csv', row.names=FALSE) Fractal Patch Code RandomFractalPatches < function(N,H,C){ # Rand omFractalPatches(N,H,C) # N = The size of array X along one dimension , # H = Hurst coeffcient; 0 < H < 1 determines fractal dimension D = 3 H , # C = The cover of the patches require("fields") A=matrix(ncol=N,nrow=N,data=0+0i) for(i in 0:N/2) {for(j in 0:N/2) {phase = 2*pi*runif(1) if((i!=0) | (j!=0)){rad=(i*i+j*j)^( (H+1)/2)*rnorm(1) } else{rad=0} A[i+1,j+1]=complex(real=rad*cos(phase), imaginary=rad*sin(phase)) if(i==0){i0=0} else{i0=N i} if(j==0){j0=0} else{j0=N j} A[i0+1,j0+1] = complex(real=rad*cos(phase),imaginary= rad*sin(phase))}} A[N/2+1,0+1]=complex(real=Re(A[N/2,0]),imaginary=0) A[0+1,N/2+1]=complex(real=Re(A[0,N/2]),imaginary=0) A[N/2+1,N/2+1]=complex(real=Re(A[N/2,N/2]),imaginary=0) for(i in 1:(N/2 1)){ for(j in 1:(N/2 1)){ phase=2*pi*runif(1) rad=(i*i+j*j)^( (H+1)/2)*rnorm(1) A[i+1,N j+1]=complex(real=rad*cos(phase),imaginary=rad*sin(phase)) A[N i+1,j+1]=complex(real=rad*cos(phase),imaginary= rad*sin(phase))}} RandomField=Re(fft(A ,inverse=TRUE)) D=sort(RandomField) Threshold=D[floor(N*N*C)+1] Result=(RandomField
PAGE 145

145 Scale=floor(1/sqrt(Cover)) NP=floor(N/(Scale)) Margin=ceiling((N Scale*(NP 1))/2) Cells=matrix(nrow=N,ncol=N,data=0) Cells[Margin+(0:(NP 1))*Scale,Margin+(0:(NP 1))*Scale]=1 cat(paste('Realized cover:', sum(Cells)/N/N)) return(Cells)}

PAGE 146

146 A PPENDIX L SUMMARY OF DROUGHT GENERATED MUDFLATS AND REMNANT PATCH COVER ACROSS THE SOUTHEASTERN US COAST T able L 1 . Cordgrass die off and resistance along the southeastern US coast. Summary of cordgrass die off extent and the distribution of cordgrass pat ches surviving in die off areas associated with mussels in marshes in Georgia and South Carolina where significan t cordgrass die off was observed in April 2012. The mean and range of measurements are shown for all mudflats observed within each marsh sites. Site (Latitude, Long) Mudflats per site Mudflat area m 2 % Mudflat vegetated Aspect ratio of mudflat All patches m 2 %Patches with mussels Measurements displayed Total Sum (range) Mean (range) Mean (range) Mean (range) Mean (range) Charleston, SC 32°46 '46.14"N, 79°57'54.07"W 5 1688 (30 759) 2.0 (0.0 3.5) 0.55 0.83 0.016 (0 0.053) 50 (17 83) Folly Beach, SC 32°46'8.30"N, 79°58'18.94"W 7 6763 (100 2880) 1.8 (0.0 6.9) 0.55 (0.36 1) 0.014 (0 0.029) 80 (23 100) West Ashley, SC 32°46'25.36"N, 80° 0'4.24"W 3 7646 (625 1351) 3.1 (1.0 7.6) 0.70 (0.31 1) 0.022 (0.014 0.031) 91 (76 100) Seabrook Isl., SC 32°34'57.07"N, 80°10'29.65"W 6 6070 (120 3750) 0.5 (0 .0 1.3) 0.33 (0.17 0.69) 0.018 (0 0.015) 63.8 (75 100) Port Royal, SC 32°23'18.16"N 80 °46'5.93"W 8 11259 (21 7350) 1.2 (0.0 0.7) 0.48 (0.16 1) 0.014 (0 0.056) 87.5 (75 100) Fort Pulaski, GA 32° 1'25.50"N, 80°55'25.35"W 11 2563 (180 435) 4.4 (4.2 4.9) 0.52 (0.17 1) 0.034 (0 0.052) 53.8 (0 100) Sapelo Isl., GA, Oakdale marsh 31°2 4'26.11"N, 81°17'24.97"W 6 1759 (44 823) 6.6 (1.4 10.4) 0.50 (0.21 0.80) 0.097 (0.023 0.17) 64.8 (9 100) Sapelo Isl., GA Lighthouse marsh 31°23'29.54"N, 81°16'32.82"W 3 2438 (151 1990) 7.7 (7.1 17.1) 0.51 (0.24 0.91) 0.161 (0.092 0.25) 86.2 (75 100) Jekyll Isl., GA 31° 5'27.68"N, 81°29'20.49" 4 11808 (90 5808) 1.2 (1.1 1.6) 0.36 (0.19 0.77) 0.046 (0.036 0.054) 67.3 (29 98)

PAGE 147

147 APPENDIX M RAINFALL IN SUMMER 2012: SALINITY MONITORING PERIOD Figure M 1. Daily precipitation on Sapelo Island, GA ove r summer 2012 when monitoring of porewater salinity was conducted (source: GCE LTER Marsh Landing weather station: http://gce ).

PAGE 148

148 LIST OF REFERENCES Alber, M., E. Swen son, S. Adamowicz, and I. A. Mendelssohn. 2008. Salt marsh dieback: an overview of recent events in the US. Estuarine, Coastal and Shelf Science 80:1 11. Altieri, A. H., and J. D. Witman. 2006. Local extinction of a foundation species in a hypoxic estuary: integrating individuals to ecosystem. Ecology 87:717 30. Altieri, A., B. Silliman, and M. Bertness. 2007. Hierarchical organization via a facilitation cascade in intertidal cordgrass bed communities. The American Naturalist 169:195 206. Altieri, A., and B . van Wesenbeeck. 2010. Facilitation cascade drives positive relationship between native biodiversity and invasion success. Ecology 91:1269 1275. Angelini, C., A. H. Altieri, B. R. Silliman, and M. D. Bertness. 2011. Interactions among Foundation Species a nd Their Consequences for Community Organization, Biodiversity, and Conservation. BioScience 61:782 789. Angelini, C., and B. Silliman. 2014. Secondary foundation species as drivers of trophic and functional diversity: evidence from a tree epiphyte system. Ecology 95:185 196. Angelini, C., and B. R. Silliman. 2012. Patch size dependent community recovery after massive disturbance. Ecology 93:101 10. Anthelme, F., and R. Michalet. 2009. Grass to tree facilitation in an arid grazed environment (Aïr Mountains, Sahara). Basic and Applied Ecology 10:437 446. Baiser, B., N. Whitaker, and A. Ellison. 2013. Modeling foundation species in food webs. Ecosphere 4:146. Barbier, E., S. Hacker, C. Kennedy, A. Stier, and B. Silliman. 2011. The value of estuarine and coasta l ecosystem services. Ecological Monographs 81:169 193. Bates, D., M. Maechler, and B. Bolker. 2012. lme4: Linear mixed effects models using S4 classes. R package version 0.999999 0. Beck, M. W., K. L. Heck, K. W. Able, D. L. Childers, D. B. Eggleston, B. M. Gillanders, B. Halpern, C. G. Hays, K. Hoshino, T. J. Minello, R. J. Orth, P. F. Sheridan, and M. P. Weinstein. 2001. The Identification, Conservation, and Management of Estuarine and Marine Nurseries for Fish and Invertebrates. BioScience 51:633.

PAGE 149

149 Belsk y, A. 1994. Influences of trees on savanna productivity: tests of shade, nutrients, and tree grass competition. Ecology 75:922 932. Benzing. 1990. Vascular Epiphytes. Pages 1 376. Cambridge University Press, Cambridge. Bertness, M. 1984. Ribbed mussels and Spartina alterniflora production in a New England salt marsh. Ecology 65:1794 1807. Bertness, M. 1985. Fiddler crab regulation of Spartina alterniflora production on a New England salt marsh. Ecology 66:1042 1055. Bertness, M., C. Crain, B. R. Silliman, M. C. Bazterrica, M. V Reyna, F. Hildago, and J. K. Farina. 2006. The community structure of western Atlantic Patagonian rocky shores. Ecological Monographs 76:439 460. Bertness, M. D., and R. Callaway. 1994. Positive interactions in communities. Trends in Ecology & Evolution 9:191 3. Bertness, M., and E. Grosholz. 1985. Population dynamics of the ribbed mussel, Geukensia demissa: the costs and benefits of an aggregated distribution. Oecologia 67:192 204. Bertrand, M., and M. Wilson. 1996. Microclimate depe ndent survival of unfed adult Ixodes scapularis (Acari: Ixodidae) in nature: Life cycle and study design implications. Journal of Medical Entomology 33:619 627. Bishop, M., J. Byers, B. Marcek, and P. Gribben. 2012. Density dependent facilitation cascades determine epifaunal community structure in temperate Australian mangroves. Ecology 93:1388 1401. Bobbink, R., K. Hicks, J. Galloway, T. Spranger, R. Alkemade, M. Ashmore, M. Bustamante, S. Cinderby, E. Davidson, F. Dentener, B. Emmett, J. Erisman, M. Fenn, F. Gilliam, A. Nordin, L. Pardo, and W. De Vries. 2010. Global assessment of nitrogen deposition effects on terrestrial plant diversity: a synthesis. Ecological Applications 20:30 59. Bracken, M., B. Bracken, and L. Rogers Bennett. 2007. Species diversity and foundation species: Potential indicators of fisheries yields and marine ecosystem functioning. CalCOFI 48:82 91. Brooker, R., F. Maestre, R. M. Callaway, C. J. Lortie, L. A. Cavieres, G. Kunstler, P. Liancourt, K. Tielborger, J. Travis, F. Anthelme, C . Armas, L. Coll, E. Corcket, S. Delzon, E. Forey, Z. Kikvidze, J. Olofsson, F. Pugnaire, C. Quirox, P. Saccone, K. Schiffers, M. Seifan, B. Touzard, and R. Michalet. 2008. Facilitation in plant communities: the past, the present, and the future. Journal o f Ecology 96:18 34.

PAGE 150

150 Bruno, J. F., and M. D. Bertness. 2001. Habitat modification and facilitation in benthic marine communities. Pages 201 216 in M. D. Bertness, S. Gaines, and M. Hay, editors. Marine Community Ecology. Sinauer. Bruno, J. F., J. J. Stachow icz, and M. D. Bertness. 2003. Inclusion of facilitation into ecological theory. Trends in Ecology & Evolution 18:119 125. Bullock, J., B. Hill, J. Silvertown, and M. Sutton. 1995. Gap colonization as a source of grassland community change: effects of gap size and grazing on the rate and mode of colonization by different species. Oikos 72:273 282. Burnham, K. P., and D. R. Anderson. 2002. Model Selection and Inference: a Practical Information Theoretical Approach. 2nd edition. Springer Verlag, New York. By ers, J. E., K. Cuddington, C. G. Jones, T. S. Talley, A. Hastings, J. G. Lambrinos, J. a Crooks, and W. G. Wilson. 2006. Using ecosystem engineers to restore ecological systems. Trends in Ecology & Evolution 21:493 500. Byers, J., P. Gribben, C. Yeager, an d E. Sotka. 2012. Impacts of an abundant introduced ecosystem engineer within mudflats of the southeastern US coast. Biological Invasions 14:2587 2600. Byrnes, J., L. Gamfeldt, F. Isbell, J. Lefcheck, J. N. Griffin, A. Hector, B. J. Cardinale, D. Hooper, L . E. Dee, and J. E. Duffy. 2014. Investigating the relationship between biodiversity and ecosystem multifunctionality: Challenges and solutions. Methods in Ecology and Evolution 5:111 124. Callaway, R. M. 1995. Positive interactions among plants. The Botan ical Review 61:306 349. Callaway, R. M., K. O. Reinhart, G. W. Moore, D. J. Moore, and S. C. Pennings. 2002. Epiphyte host preferences and host traits: mechanisms for species specific interactions. Oecologia 132:221 230. Canion, C., and K. Heck. 2009. Effe ct of habitat complexity on predation success: re evaluating the current paradigm in seagrass beds. Marine Ecology Progress Series 393:37 46. Cardinale, B., J. Duffy, A. Gonzalez, D. Hooper, C. Perrings, P. Venail, A. Narwani, G. Mace, D. Tilman, D. Wardle , A. Kinzig, G. Daily, M. Loreau, J. Grace, A. Larigauderie, D. Srivastava, and S. Naeem. 2012. Biodiversity loss and its impact on humanity. Nature 486:59 67. Cardinale, B., K. Matulich, D. U. Hooper, J. E. Duffy, L. Gamfeldt, P. Balvanera, M. I. , and A. Gonzalez. 2011. The functional role of producer diversity in ecosystems. American Journal of Botany 98:572 92.

PAGE 151

151 Carney, L., A. Bohonak, M. Edwards, and F. Alberto. 2013. Genetic and experimental evidence for a mixed age, mixed origin bank of kelp m icroscopic stages in southern California. Ecology 94:1955 1965. Coleman, F., and S. Williams. 2002. Overexploiting marine ecosystem engineers: potential consequences for biodiversity. Trends in Ecology & Evolution 17:40 44. Connell, J. 1978. Diversity in t ropical rain forests and coral reefs. Science 199:1302 1310. Craft, C., E. Seneca, and S. Broome. 1991. Loss on ignition and Kjeldahl digestion for estimating organic carbon and total nitrogen in estuarine marsh soils: calibration with dry combustion. Estu aries 14:175 179. Crain, C. M., and M. Bertness. 2006. Ecosystem engineering across environmental gradients: implications for conservation and management. Bioscience 56:211 218. Crowder, L. B., and W. E. Cooper. 1982. Habitat structural complexity and the interaction between bluegills and their prey. Ecology 63:1802 1813. Cruz Angón, A., M. L. Baena, and R. Greenberg. 2009. The contribution of epiphytes to the abundance and species richness of canopy insects in a Mexican coffee plantation. Journal of Tropic al Ecology 25:453 463. Currin, C., S. Newell, and H. Paerl. 1995. The role of standing dead Spartina alterniflora and benthic microalgae in salt marsh food webs: considerations based on multiple stable isotope analysis. Marine Ecology Progress Series 121:9 9 116. Dayton, P. K. 1972. Toward an understanding of community resilience and the potential effects of enrichments to the benthos at McMurdo Sound, Antarctica. Proceedings of the colloquium on conservation problems in Antarctica. Allen Press, Lawrence, Ka nsas. Dial, R., M. Ellwood, and E. Turner. 2006. Arthropod abundance, canopy structure, and microclimate in a Bornean lowland tropical rain forest. Biotropica 38:643 652. Dijkstra, J. A., J. Boudreau, and M. Dionne. 2012. Species specific mediation of temp erature and community interactions by multiple foundation species. Oikos 121:646 654. Duarte, C. M. 2000. Marine biodiversity and ecosystem services: an elusive link. Journal of Experimental Marine Biology and Ecology 250:117 131.

PAGE 152

152 Ellison, A., M. Bank, B. D. Clinton, E. A. Colburn, K. Elliot, C. R. Ford, D. R. Foster, B. D. Kloeppel, J. D. Knoepp, G. M. Lovett, J. Mohan, D. A. Orwig, N. L. Rodenhouse, W. V Sobezak, K. A. Stinson, J. K. Stone, C. M. Swan, J. Thompson, B. Von Holle, and J. R. Webster. 2005. L oss of foundation species: consequences for the structure and dynamics of forested ecosystems. Frontiers in Ecology 3:479 486. Ellwood, M. D. F., and W. A. Foster. 2004. Doubling the estimate of invertebrate biomass in a rainforest canopy. Nature 429:549 5 51. Ellwood, M. D. F., R. G. W. Northfield, M. Mejia Chang, and H. Griffiths. 2011. On the vapour trail of an atmospheric imprint in insects. Biology Letters 7:601 4. Faunce, C. H., and J. E. Serafy. 2008. Growth and secondary production of an eventual ree f fish during mangrove residency. Estuarine, Coastal and Shelf Science 79:93 100. Folke, C., S. Carpenter, B. Walker, T. Elmqvist, L. Gunderson, and C. Holling. 2004. Regime shifts, resilience, and biodiversity in ecosystem management. Annual Review of Eco logy, Evolution, and Systematics 35:557 581. Franco, A., and P. Nobel. 1989. Effect of nurse plants on the microhabitat and growth of cacti. Journal of Ecology 77:870 886. Franks, S. 2003. Facilitation in multiple life history stages: evidence for nucleate d succession in coastal dunes. Plant Ecology 168:273 282. Freiberg, M. 2001. The influence of epiphyte cover on branch temperature in a tropical tree. Plant Ecology 153:241 250. Gamfeldt, L., H. Hillebrand, and P. Jonsson. 2008. Multiple functions increase the importance of biodiversity for overall ecosystem functioning. Ecology 89:1223 1231. Gómez Aparicio, L. 2009. The role of plant interactions in the restoration of degraded ecosystems: a meta analysis across life forms and ecosystems. Journal of Ecology 97:1202 1214. Goodsell, P. J., M. J. Fowler Walker, B. M. Gillanders, and S. D. Connell. 2004. Variations in the configuration of algae in subtidal forests: Implications for invertebrate assemblages. Austral Ecology 29:350 357. Gotelli, N. J., and R. K. C olwell. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters 4:379 391.

PAGE 153

153 Gribben, P. E., J. E. Byers, M. Clements, L. a McKenzie, P. D. Steinberg, and J. T. Wright. 2009. Behavioural interactions between ecosystem engineers control community species richness. Ecology Letters 12:1127 36. Guichard, F., P. M. Halpin, G. W. Allison, J. Lubchenco, and B. A. Menge. 2003. Mussel The American Naturalist 161:889 904. Halpern, B. S., B. Silliman, J. D. Olden, J. F. Bruno, and M. D. Bertness. 2007. Incorporating positive interactions in aquatic restoration and conservation. Frontiers in Ecology and the Environment 5:153 160. Harvey, J., and W. Nuttle. 1995. Fluxes of water and solute in a coastal wetland sediment. 2. Effect of macropores on solute exchange with surface water. Journal of Hydrology 164:109 125. Hay, M. 1986. Associational plant defenses and the maintenance of species diversity: turning competitors into accomplices. American Naturalist 128:617 641. He, Q. , M. Bertness, and A. Altieri. 2013. Global shifts towards positive species interactions with increasing environmental stress. Ecology Letters 16:695 706. Heck, K., G. Hays, and R. Orth. 2003. Critical evaluation of the nursery role hypothesis for seagras s meadows. Marine Ecology Progress Series 253:123 136. Hector, A., and R. Bagchi. 2007. Biodiversity and ecosystem multifunctionality. Nature 448:05947. Hemond, H., and W. Nuttle. 1984. Surface infiltration in salt marshes: Theory, measurement, and biogeoc hemical implications. Water Resources Research 20:591 600. Hensel, M., and B. Silliman. 2013. Consumer diversity across kingdoms supports multiple functions in a coastal ecosystem. Proceedings of the National Academy of Sciences of the United States of Ame rica 110:20621 20626. Hensel, P., D. Pont, and J. Day. 1998. Short term sedimentation dynamics in the Rhône River Delta. France: The importance of riverine pulsing. Estuaries 21:52 65. Hooper, D., E. Adair, B. J. Cardinale, J. E. K. Byrnes, B. Hungate, K. L. Matulich, A. reveals biodiversity loss as a major driver of ecosystem change. Nature 486:105 108.

PAGE 154

154 Hooper, D., F. I. Chapin, J. Ewel, A. Hector, P. Inchausti, S. Lavorel, J. Lawton, D. Lodge, M. Loreau, S. Naeem, B. Schmid, H. Setala, A. Symstad, J. Vandermeer, and D. Wardle. 2005. Effects of biodiversity on ecosystem functioning: a consensus of current knowledge. Ecological Monographs 75:3 35. Hsieh, C. F., Z. S. Chen, Y. M. Hsu, K. C. Yang, and T. H. Hsieh. 2009. Altitudinal zonation of evergreen broad leaved forest in Mount Lopei, Taiwan. Journal of Vegetation Science 9:201 212. Hughes, A. R., J. E. Byrnes, D. L. Kimbro, and J. J. Stachowicz. 2007. Reciprocal relationships and potential feedbacks between biodiversity and disturbance. Ecology Letters 10:849 64. Hughes, A. R., and J. J. Stachowicz. 2004. Genetic diversity enhances the resistance of a seagrass ecosystem to disturbance. Proceedings of the National Academy of Sci ences of the United States of America 101:8998 9002. Hughes, A., A. Wilson, and J. Morris. 2012. Hydrologic variability in a salt marsh: Assessing the links between drought and acute marsh dieback. Estuarine, Coastal and Shelf Science 111:95 106. IPCC. 201 4. Intergovernmental Panel on Climate Change. United Nations. Irlandi, E. A. 1994. Large and small scale effects of habitat structure on rates of predation: how percent coverage of seagrass affects rates of predation and siphon nipping on an infaunal biva lve. Oecologia 98:176 183. Irving, A. D., and M. D. Bertness. 2009. Trait dependent modification of facilitation on cobble beaches. Ecology 90:3042 50. Irving, A. D., S. D. Connell, and B. M. Gillanders. 2004. Local complexity in patterns of canopy benthos associations produces regional patterns across temperate Australasia. Marine Biology 144:361 368. Jaschinski, S., N. Aberle, S. Gohse Reimann, H. Brendelberger, K. H. Wiltshire, and U. Sommer. 2009. Grazer diversity effects in an eelgrass epiphyte microph ytobenthos system. Oecologia 159:607 15. Jost, J., and B. Helmuth. 2007. Morphological and ecological determinants of body temperature of Geukensia demissa, the Atlantic ribbed mussel, and their effects on mussel mortality. The Biological Bulletin. Jost, L . 2006. Entropy and diversity. Oikos 113:363 375.

PAGE 155

155 Kappelle, M., P. A. Kennis, and R. A. Vries. 1995. Changes in diversity along a successional gradient in a Costa Rican upper montane Quercus forest. Biodiversity and Conservation 4:10 34. Kiers, E., T. Palm er, A. Ives, J. Bruno, and J. Bronstein. 2010. Mutualims in a changing world: an evolutionary perspective. Ecology Letters 13:1459 1474. for community structure Annual Re view of Entomology 46: 729 760. Annual Review of Entomology 46:729 760. Knowlton, N., and J. Jackson. 2001. The ecology of coral reefs. Pages 395 422 in M. D. Bertness, S. Gaines, and M. E. Hay, editors. Marine Community Ecology. Sinauer. Koch, E. W., E. B . Barbier, B. R. Silliman, D. J. Reed, G. M. Perillo, S. D. Hacker, E. F. Granek, J. H. Primavera, N. Muthiga, S. Polasky, B. S. Halpern, C. J. Kennedy, C. V Kappel, and E. Wolanski. 2009. Non linearity in ecosystem services: temporal and spatial variabili ty in coastal protection. Frontiers in Ecology and the Environment 7:29 37. Van de Koppel, J., D. van der Wal, J. Bakker, and P. M. J. Herman. 2005. Self organization and vegetation collapse in salt marsh ecosystems. The American Naturalist 165:E1 E12. Kue nzler, E. 1961. Structure and energy flow of a mussel population in a Georgia salt marsh. Limnology and Oceanography 6:191 204. Laegdsgaard, P., and C. Johnson. 2001. Why do juvenile fish utilise mangrove habitats? Journal of Experimental Marine Biology a nd Ecology 257:229 253. Lamb, D., P. Erskine, and J. Parrotta. 2005. Restoration of degraded tropical forest landscapes. Science 310:1628 1632. Letzch, W., and R. Frey. 1980. Deposition and erosion in a Holocene salt marsh, Sapelo Island, GA. Journal of Se diment Petrology 50:529 542. Levin, S., and R. Paine. 1974. Disturbance, patch formation, and community structure. Proceedings of the National Academy of Sciences of the United States of America 71:2744 2747. Levine, J. M., J. S. Brewer, and M. D. Bertness . 1998. Nutrients, competition and plant zonation in a New England salt marsh. Ecology 86:285 292. Lotze, H., H. Lenihan, and B. Bourque. 2006. Depletion, degradation, and recovery potential of estuaries and coastal seas. Science 312:1806 1809.

PAGE 156

156 Ludwig, F., T. Dawson, and H. Prins. 2004. Below ground competition between trees and grasses may overwhelm the facilitative effects of hydraulic lift. Ecology Letters 7:623 631. MacArthur, R. 1965. Patterns in species diversity. Biological Reviews 40:510 533. Maestr e, F. T., R. M. Callaway, F. Valladares, and C. J. Lortie. 2009. Refining the stress gradient hypothesis for competition and facilitation in plant communities. Journal of Ecology 97:199 205. Marion, S. R., and R. J. Orth. 2008. Innovative Techniques for La rge scale Seagrass Restoration Using Zostera marina (eelgrass) Seeds. Restoration Ecology 18:514 526. Martin, C. E., and J. N. Siedow. 1981. Crassulacean Acid Metabolism in the epiphyte Tillandsia usneoides L. (Spanish Moss). Plant Physiology 68:335 339. M atelson, T. J., N. M. Nadkarni, and J. Longino. 1993. Longevity of fallen epiphytes in a Neotropical montane forest. Ecology 74:265 269. Mazerolle, M. J. 2013. Model selection and multimodel inference based on (Q)AIC(c. McAuliffe, J. 1984. Sahuaro nurse tr ee associations in the Sonoran Desert: competitive effects of sahuaros. Oecologia:695 706. McKee, K. L., I. A. Mendelssohn, and M. D. Materne. 2004. Acute salt marsh dieback in the Mississippi River detaic plain: a drought induced phenomenon? Global Ecolo gy and Biogeography 13:65 73. Mendelssohn, I. a., and E. D. Seneca. 1980. The influence of soil drainage on the growth of salt marsh cordgrass Spartina alterniflora in North Carolina. Estuarine and Coastal Marine Science 11:27 40. Mendelssohn, I., and J. M orris. 2010. Eco physiological controls on the productivity of Spartina Alterniflora Loisel. in A. Kreeger, editor. Concepts and Controversies in Tidal Marsh Ecology. Kluwer Academic, Dordrecht. Menge, B. A., and J. P. Sutherland. 1976. Species diversity g radients: synthesis of the roles of predation, competition, and temporal heterogeneity. American Naturalist 110:351 369. Menge, B., and J. Sutherland. 1987. Community regulation: variation in disturbance, competition, and predation in relation to environme ntal stress and recruitment. American Naturalist.

PAGE 157

157 Miller, D., R. Geider, and H. MacIntyre. 1996. Microphytobenthos: the ecological role of water marine habitats. II. Role in sediment stability and shallow water f ood webs. Estuaries 19:202 212. Mondragn, D., L. M. Calvo Irabien, and D. H. Benzing. 2004. The basis for obligate epiphytism in Tillandsia brachycaulos (Bromeliaceae) in a Mexican tropical dry forest. Journal of Tropical Ecology 20:97 104. Montague, C. 19 80. A natural history of temperate western Atlantic fiddler crabs (Genus Uca) with reference to their impact on the salt marsh. Contributions in Marine Science 23:25 55. Nadkarni, N. 1989. Bird use of epiphyte resources in neotropical trees. Condor 91:891 907. Nadkarni, N. 1992. Biomass and nutrient dynamics of epiphytic litterfall in a neotropical montane forest, Costa Rica. Biotropica 24:24 30. Nadkarni, N. 1994. Diversity of species and interactions in the upper tree canopy of forest ecosystems. American Zoologist 34:70 78. Orth, R., T. Carruthers, and W. Dennison. 2006. A global crisis for seagrass ecosystems. Bioscience 56:987 996. Orth, R., K. Heck, and J. van Montfrans. 1984. Faunal communities in seagrass beds: a review of the influence of plant stru cture and prey characteristics on predator prey relationships. Estuaries 7:339 350. Paine, R., and S. Levin. 1981. Intertidal landscapes: disturbance and the dynamics of pattern. Ecological Monographs 51:145 178. Palomo, L., C. Meile, and S. Joye. 2013. Dr ought impacts on biogeochemistry and microbial processes in salt marsh sediments: a flow through reactor approach. Biogeochemistry 112:389 407. Pandolfi, J., S. Connolly, D. Marshall, and A. Cohen. 2011. Projecting coral reef futures under global warming a nd ocean acidification. Science 333:418 422. Pasari, J., T. Levi, E. Zavaleta, and D. Tilman. 2013. Several scales of biodiversity affect ecosystem multifunctionality. Proceedings of the National Academy of Sciences of the United States of America 110:1021 9 10222. Pascual, M., and F. Guichard. 2005. Criticality and disturbance in spatial ecological systems. Trends in Ecology & Evolution 20:88 95.

PAGE 158

158 Pennings, S. C., M. Grant, and M. Bertness. 2005. Plant zonation in low latitude salt marshes: disentangling the roles of flooding, salinity and competition. Journal of Ecology 93:159 167. Pennings, S. C., and C. Richards. 1998. Effects of wrack burial in salt stressed habitats: Batis maritima in a southwest Atlantic salt mar sh. Ecography 21:630 638. Persson, L. 1995. Prey refuges affecting interactions between piscivorous perch and juvenile perch and roach. Ecology 76:70 81. Pringle, R., D. Doak, A. Brody, R. Joque, and T. Palmer. 2010. Spatial pattern enhances ecosystem func tion. PLoS Biology 8:1000377. R Core Development Team. 2012. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Riginos, C. 2009. Grass competition suppresses savanna tree growth across multipl e demographic stages. Ecology 90:335 40. Ripley, B. 1981. Spatial statistics. Pages 1 252. Wiley, Hoboken, New Jersey. Rossi, J. P. 2011. rich: An R Package to Analyse Species Richness. Pages 112 120. Rowntree, J. K., D. M. Shuker, and R. F. Preziosi. 2011 . Forward from the crossroads of ecology and evolution. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 366:1322 8. Schalles, J., C. Hladik, A. Lynes, and S. C. Pennings. 2013. Landscape estimates of habitat types, plant biomass, and invertebrate densities in a Georgia salt marsh. Oceanography 26:88 98. Silliman, B., J. van de Koppel, M. D. Bertness, L. E. Stanton, and I. A. Mendelssohn. 2005. Drought, snails, and large scale die off of southern US salt marshes. Scie nce 310:1803 1806. Silliman, B., C. Layman, K. Geyer, and J. Zieman. 2004. Predation by the black clawed mud crab, Panopeus herbstii, in mid Atlantic salt marshes: further evidence for top down control of marsh grass production. Estuaries 27:188 196. Silli man, B. R., M. D. Bertness, A. H. Altieri, J. N. Griffin, M. C. Bazterrica, F. J. Hidalgo, C. M. Crain, and M. V Reyna. 2011. Whole community facilitation regulates biodiversity on Patagonian rocky shores. PLoS ONE 6:e24502. Silliman, B. R., M. McCoy, C. A ngelini, R. D. Holt, J. N. Griffin, and J. van de Koppel. 2013. Consumer fronts, climate change, and runaway collapse of ecosystems. Annual Review of Ecology, Evolution, and Systematics 44.

PAGE 159

159 Silliman, B. R., and S. Newell. 2003. Fungal farming in a snail. P roceedings of the National Academy of Sciences of the United States of America 100:15643 15648. Silliman, B., and J. Zieman. 2001. Top down control of Spartina alterniflora production by periwinkle grazing in a Virginia salt marsh. Ecology 82:2830 2845. Si mpson, J., E. Slade, T. Riutta, and M. Taylor. 2012. Factors affecting soil fauna feeding activity in a fragmented lowland temperate deciduous woodland. PLoS ONE 7:e29616. Smith, J., and R. Frey. 1985. Biodeposition by the ribbed mussel Geukensia demissa i n a salt marsh, Sapelo Island, Georgia. Journal of Sedimentary Research 55:817 828. Smith, L., and T. Hughes. 1999. An experimental assessment of survival, re attachment and fecundity of coral fragments. Journal of Experimental Marine Biology and Ecology 2 35:147 164. Solorzano, L. 1969. Determination of ammonia in natural waters by the phenolhypochlorite method. Limnology and Oceanography 14:799 801. Soomdat, N. N., M. McCoy, M. Hensel, S. Buhler, Z. A. Chejanovski, and B. R. Silliman. 2014. Independent and combined effects of multiple predators across ontogeny of a dominant grazer. Oikos. Sousa, W. 1984. The role of disturbance in natural communities. Annual Review of Ecology, Evolution, and Systematics 15:353 391. Sousa, W. 2001. Natural disturbance and th e dynamics of marine benthic communities. Pages 85 130 in M. Bertness and S. Gaines, editors. Marine Community Ecology. 1st edition. Sinauer. Sousa, W., and P. Kennedy. 2007. Supply side ecology in mangroves: do propagule dispersal and seedling establishme nt explain forest structure? Ecological Monographs 77:53 76. Stachowicz, J. J. 2001. Mutualism, Facilitation, and the Structure of Ecological Communities. BioScience 51:235. Stiven, A., and S. Gardner. 1992. Population processes in the ribbed mussel Geuken sia demissa(Dillwyn) in a North Carolina salt marsh tidal gradient: spatial pattern, predation, growth and. Journal of Experimental Marine Biology and Ecology 160:81 102.

PAGE 160

160 Stuntz, S., U. Simon, and G. Zotz. 2002. Rainforest air conditioning: the moderating influence of epiphytes on the microclimate in tropical tree crowns. International Journal of Biometeorology 46:53 59. Syndonia Bret Harte, M., M. Mack, G. Shaver, D. Huebner, M. Johnston, C. Mojica, C. Pizano, and J. Reiskind. 2013. The response of Arctic vegetation and soils following an ususually severe tundra fire. Philosophical transactions of the Royal Society of London. Series B, Biological sciences 368:20120490. Thomsen, M. S., T. Wernberg, A. Altieri, F. Tuya, D. Gulbransen, K. J. McGlathery, M. Hol mer, and B. R. Silliman. 2013. Habitat cascades: the conceptual context and global relevance of facilitation cascades via habitat formation and modification. Integrative and Comparative Biology 50:158 75. Travis, S., and M. Hester. 2005. A space for time s ubstitution reveals the long term decline in genotypic diversity of a widespread salt marsh plant, Spartina alterni flora, over a span of 1500 years. Journal of Ecology 93:417 430. Tscharntke, T., and A. Klein. 2005. Landscape perspectives on agricultural intensification and biodiversity ecosystem service management. Ecology Letters 8:857 874. Turner, R. M., S. M. Alcorn, and J. A. Booth. 1966. The influence of shade, soil, and water on saguaro seedling establishment. Botanical Gazette 127:95 102. Valentin e, J. F., and K. L. Heck. 1993. Mussels in seagrass meadows: their influence on macroinvertebrate abundance and secondary production in the northern Gulf of Mexico. Marine Ecology Progress Series 96:63 74. Venables, W., and B. Ripley. 2002. Modern Applied Statistics with S. 4th edition. Springer. Wesenbeeck, B., C. Crain, A. H. Altieri, and M. Bertness. 2007. Distinct habitat types arise along a continuous hydrodynamic stress gradient due to interplay of competition and facilitation. Marine Ecology Progress Series 349:695 706. Whitham, T. G., J. K. Bailey, J. a Schweitzer, S. M. Shuster, R. K. Bangert, C. J. LeRoy, E. V Lonsdorf, G. J. Allan, S. P. DiFazio, B. M. Potts, D. G. Fischer, C. a Gehring, R. L. Lindroth, J. C. Marks, S. C. Hart, G. M. Wimp, and S. C. Wooley. 2006. A framework for community and ecosystem genetics: from genes to ecosystems. Nature reviews. Genetics 7:510 23. Wigglesworth, V. 1945. Transpiration through the cuticle of insects. Journal of Experimental Biology:97 115.

PAGE 161

161 Wolf, B., and C. M artinez del Rio. 2003. How important are columnar cacti as sources of water and nutrients for desert consumers? A review. Isotopes in Environmental and Health Studies 39:53 67. Wulff, J. 1984. Sponge mediated coral reef growth and rejuvenation. Coral Reefs 3:157 163. Yakovis, E., A. Artemieva, N. Shunatova, and M. Varfolomeeva. 2008a. Multiple foundation species shape benthic habitat islands. Oecologia 155:785 795. Yakovis, E. L., A. V. Artemieva, N. N. Shunatova, and M. A. Varfolomeeva. 2008b. Multiple fou ndation species shape benthic habitat islands. Oecologia 155:785 95. Yanoviak, S. P., S. M. Berghoff, K. E. Linsenmair, and G. Zotz. 2011. Effects of an epiphytic orchid on arboreal ant community structure in Panama. Biotropica 43:731 737. Yarranton, G., a nd R. Morrison. 1974. Spatial dynamics of a primary succession: nucleation. Journal of Ecology 62:417 428. Zahawi, R., and K. Holl. 2013. Testing applied nucleation as a strategy to facilitate tropical forest recovery. Journal of Applied Ecology 50:88 96. Zavaleta, E., J. Pasari, K. Hulvey, and D. Tilman. 2010. Sustaining multiple ecosystem functions in grassland communities requires higher biodiversity. Proceedings of the National Academy of Sciences of the United States of America 107:1443 1446. Zotz, G., and P. Hietz. 2001. The physiological ecology of vascular epiphytes: current knowledge, open questions. Journal of Experimental Botany 52:2067 78. Zytynska, S. E., M. F. Fay, D. Penney, and R. F. Preziosi. 2011. Genetic variation in a tropical tree specie s influences the associated epiphytic plant and invertebrate communities in a complex forest ecosystem. Philosophical transactions of the Royal Society of London. Series B, Biological Sciences 366:1329 36.

PAGE 162

162 BIOGRAPHICAL SKETCH Christine Angelini graduate d cum laude from the Hotchkiss School, Lakeville, CT in 2003. She then received her Bachelor of Science with High Honors in marine biology from Brown University in Providence, RI in 2007. After finishing her undergraduate studies, she spent two years worki ng under Mark Bertness at Brown University as a research technician and lab manager. In 2014, she received her PhD in Zoology from the Department of Biology at the University of Florida. Upon completion of her doctoral degree, she transitioned into an assi stant professor position in the Department of Environmental Engineering at the University of Florida in Gainesville, Florida in August 2014.