1 RESTORING PATTERN WITHOUT PROCESS IN LAKE RESTORATION: A LARGESCALE LITTORAL HABITAT ENHANCEMEN T PROJECT ON LAKE TOHOPEKALIGA, FLORIDA By ZACHARIAH C. WELCH A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORID A IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
2 Zachariah C. Welch
3 ACKNOWLEDGMENTS My experiences at the Fl orida Cooperative Research Unit, where I worked and studied from 1999 to 2009, far exceeded my most optimistic expectations. My advisor, Wiley Kitchens, kept me involved in a mu ltitude of projects and tasks, inundating me with anything and everything to do with re search and academia. The environment he and Franklin Percival provide for student s and staff is unparalleled and directly responsible for any accomplishments I have had over the last 10 years. Their unique combination of humor, guidance, support, freedom, and pressure to perform beyond your own expectations is truly a recipe for success. This is clearly evident by the caliber of my peers, all of whom were instrum ental at different stages of my graduate experience. I would also like to thank Pe ter Frederick, whose undergraduate Wetlands Ecology course I took so many years ago inspired me to pursue all areas wet and muddy, introducing me to the greatest ecosystems on the planet. I would be remiss not to thank the people who gave me the strength to tackle all of lifes challenges with unwavering love, s upport and guidance: my parents, Curt and Sandy, and my wife, Christa Zweig. My paren ts are ultimately responsible for all my past and future successes, having instilled in me every good quality I possess. My wife, whom I met during my graduate career, was an ear when I had to vent, a colleague when I needed sound advice, a shoulder if I had to lean, and a refuge when I wanted to get away. Thanks to all those who support ed me through this wonderful experience.
4 TABLE OF CONTENTS page ACKNOWLEDGMENTS ..............................................................................................................3 LIST OF TABLES .........................................................................................................................6 LIST OF FIGURES .......................................................................................................................7 ABSTRACT ............................................................................................................................... ..10 CHAPTER 1 INTRODUCTION .................................................................................................................12 Ecosystem Restoration ......................................................................................................12 Lake Tohopekaliga ..............................................................................................................15 Brief History ..................................................................................................................17 Previous Studies ..........................................................................................................19 Study Objectives and Methodology ..................................................................................20 Pattern without Process ..............................................................................................21 Mechanical Muck Removal Effects ...........................................................................22 Effects of Environmental Disturbances on Muck Removal Project ......................23 2 RESTORING LITTORAL VEGETATION PATTERNS WITHOUT KEY STRUCTURING PROCESSES ........................................................................................24 Introduction ..........................................................................................................................24 Methods and Analyses .......................................................................................................28 Sample Locations ........................................................................................................30 Analyses ........................................................................................................................33 Pre/post restoration comparison ........................................................................33 Temporal analyses ...............................................................................................37 Results ............................................................................................................................... ...37 Hydrological Comparisons to Historical Record .....................................................37 Vegetation Response ..................................................................................................40 Discussion ............................................................................................................................51 Historical Comparisons ...............................................................................................51 Management Implications ...........................................................................................55 3 MUCK REMOVAL AS A TOOL FOR RESTORATION OF LITTORAL VEGETATION ON A SUBTROPICAL LAKE ..................................................................60 Introduction ..........................................................................................................................60 Methods and Analyses .......................................................................................................63
5 Sample Locations ........................................................................................................64 Analyses ........................................................................................................................68 Pre/post-restoration comparison ........................................................................68 Temporal analyses ...............................................................................................71 Results ............................................................................................................................... ...71 Soils ...............................................................................................................................71 Vegetation .....................................................................................................................72 Time Series ...................................................................................................................74 Community Changes ...................................................................................................77 Discussion ............................................................................................................................82 Muck Removal ..............................................................................................................82 Management Implications ...........................................................................................85 4 HURRICANE EFFECTS ON A LA RGE-SCALE LITTO RAL HABITAT RESTORATION PROJECT ...............................................................................................89 Introduction ..........................................................................................................................89 Methods ............................................................................................................................... .91 Results ............................................................................................................................... ...93 Discussion ..........................................................................................................................106 5 DISCUSSION ....................................................................................................................113 Review ............................................................................................................................... .113 Management Implications ................................................................................................117 APPENDIX SPECIES LIST ..........................................................................................................................121 LIST OF REFERENCES .........................................................................................................123 BIOGRAPHICAL SKETCH .....................................................................................................133
6 LIST OF TABLES Table page 2-1 Indicator values (0) of species in the six groups identified by the cluster analysis. ...........................................................................................................................42 3-1 Change in total dry biomass (g) from two preand two post-treatment sample periods of common species. .........................................................................................74 3-2 Indicator values of species in the pre-enhancement period (Dec 2002, Jun 2003), with values ranging 0. ..............................................................................78 3-3 Indicator values of species in the post-enhancement period (Dec 2007, Jun 2008), with values ranging 0. ..............................................................................79 A Common species sampled over the period of study. .............................................121
7 LIST OF FIGURES Figure page 1-1 Location of Lake Toho pekaliga and East Lake Toho in relation to Lake Okeechobee and the Kissimmee River. .....................................................................16 1-2 Daily mean water elevations in me ters (NGVD) from January 1942 until Jun 2008. ............................................................................................................................... ..18 2-1 Location of study sites throughout the lake. ..............................................................31 2-2 Monthly lake stage (Meters NGVD) for the pre-regulat ion period 1942 (green), and post-regulat ion period 1965 (red). .............................................39 2-3 Daily lake stage (Meters NGVD) fo r a pre-regulation (1944) and postregulation per iod (1992). ....................................................................................40 2-4 Frequency histogram displaying number of species per quadrat for all samples before and after restoration. .........................................................................41 2-5 Frequency histogram of species gain or loss by quadrat after restoration. ..........41 2-6 CART model of pre-post restoration periods (9 groups, CV error = 0.36, Misclass rate = 17%). ....................................................................................................44 2-7 NMS ordination of two preand tw o post-restoration sample periods. Approximate community boundaries based on cluster groupings were outlined on the ordination fo r ease of inte rpretation................................................ 46 2-8 NMS ordination of two preand tw o post-restoration sample periods. Approximate community boundaries based on cluster groupings were overlayed onto the ordinat ion, and arrows show movement of samples through species spac e over time............................................................................ 47 2-9 MRT model of pre-post restoration periods (9 groups, CV error = 0.55). .............49 2-10 Changes in biomass of common specie s (in kilograms wet weight), including Nuphar advena and Nymphaea odorata grouped as lily pads. ..............................51 3-1 Location of experimenta l study sites on a bathymetric map of Lake Toho (0 13 ft, or ~ 2 m). ...............................................................................................................66 3-2 A boxplot of percent organic content in all soil cores from treatment and control plots, before (2002) and afte r (2008) the dry-down and muck removal project (before aver aging sub-samples). ....................................................................72
8 3-3 Species richness by depth category and treatment, before and after restoration. .......................................................................................................................73 3-4 NMS ordinations of each site, wit h depth categories graphically separated. .......76 3-5 CART model of pre-treatment comm unity distribution along the measured environmental gradients. ...............................................................................................80 3-6 CART model of post-treatment co mmunity distribution along the measured environmental gradients. ...............................................................................................82 4-1 Lake stages (meters above sea le vel NGVD) following the managed dry down in summer of 2004. ..............................................................................................92 4-2 Initial relative abundances of dom inant species in scraped experimental plots. ............................................................................................................................... ..94 4-3 Initial relative abundances of dom inant species in control (unscraped) experimental plots. .........................................................................................................95 4-4 Initial species compositions of domi nant species in scraped sections of lakewide study sites. .............................................................................................................96 4-5 Average wet biomass per site (kg) of all species in the shallow, scraped (gray) and deeper water, unscraped (black) sections of the lake-wide study sites. ............................................................................................................................... ..97 4-6 Average wet biomass per site of A) P. acuminatum (grams) in the shallow, scraped (gray) and deeper water, unscraped (black) sections of the lakewide study sites, and B) P. geminatum (kilograms). ................................................98 4-7 Average dry biomass per site (g) of P. acuminatum in control (gray) and treated (black) sections of experimental plots. ..........................................................99 4-8 Number of empty samples in experim ental plots (gray and black) and lakewide study sites. ...........................................................................................................101 4-9 Average daily lake stage (meters NGVD) with horizontal dashed lines representing normal maximum and minimum lake stages. ...................................102 4-10 Average wet biomass (g) of the five dom inant species at the end of the study in the lake-wide sites. ..................................................................................................103 4-11 Average dry biomass (g) of the five dominant specie s at the end of the study in the A) treated plots and B) control plots of the experimental study sites. .......104
9 4-12 Average number of Vallisneria plants (per m2) across all samples, including experimental and lake-wide study sites. ...................................................................105
10 Abstract of Dissertation Pr esented to the Graduate School of the University of Florida in Partial Fulf illment of the Requirements for t he Degree of Doctor of Philosophy RESTORING PATTERN WITHOUT PROCESS IN LAKE RESTORATION: A LARGESCALE LITTORAL HABITAT MODIFICAT ION PROJECT ON LAKE TOHOPEKALIGA, FLORIDA By Zachariah C. Welch December 2009 Chair: Wiley M. Kitchens Major: Interdisciplinary Ecology Muck removal is an extreme habitat restor ation technique, where heavy machinery is used to remove dense vegetation and a ccumulated organic material from lake shorelines. This approach is used to combat accelerated plant growth and subsequent litter deposition, following decades of altere d hydrologic schedules, elevated nutrient levels, and exotic species introductions. This study explored the efficacy of the technique, using results from the larges t muck-removal application to date. Vegetation communities were monitored from 2002 to 2008, including two years prior to restoration and four year s after. Before muck removal, Pontederia cordata represented the dominant community in the z one of annual water level fluctuation, as compared to Panicum repens which was dominant in the 1950s. Four years after muck removal, P. repens had expanded slightly from its pr e-restoration levels, but the dominant species in treated areas was Vallisneria americana This submersed species had no record of ever being prevalent in the system, and represented a novel community.
11 Muck removal effects were short-lived in t he shallowest areas of the littoral zone, with sites < 0.75 meters in depth recovering to pre-treatment levels within 3 years of reflooding. Deeper sites (> 1.0 meter) gener ally had the largest impacts, with entirely new communities established within three y ears of treatment. These results suggest that future projects focu s on deeper-water emergent comm unities, as dense vegetation may quickly recolonize shallow shorelines. There was concern among managers t hat hurricanes passing over the lake immediately following muck removal dramatically altered the outcome of the project, but this study found otherwise. While co rresponding high water events undoubtedly delayed recovery in treated areas, the com positions of early colonizers were not changed. The same compositions found just prior to hurricane passage were found the following growing season, and the eventually -dominant communities did not appear until nearly two years after the stor m events. These results suggest that initial water levels were less important in determining restor ed community types than the longer-term hydroperiods within treated areas.
12 CHAPTER 1 INTRODUCTION Ecosystem Restoration Ecosystem restoration has become increas ingly important over the last decade (Suding et al. 2004) and its application increasi ngly complex. In the simplest case restoration and degradat ion can be depicted as linear proc esses traveling in opposite directions along parallel pathways (Dobson et al. 1997), with recovery occurring when the appropriate stressor is removed from the system. Restoration e fforts usually begin by re-establishing historical structure or s pecies mix (pattern), and/or the environmental conditions (processes) that permit a site to become self sustaining (Parker 1997). For example, prairie restorati on may begin with re-introducing fire to eliminate woody species encroachment (Doren and Whiteak er 1990). This "applied succession" approach (Niering 1987), where re-establishi ng key structuring processes produces a desired pattern through self-organization (Mi tsch and Wilson 1996) works well when a single constraint exists (Suding et al. 2004). The degree to which restoration is possibl e, however, is generally limited by the severity of degradation and the effort that can be applied (National Research Council 1992). In reality, degradation is not a linear process and involves multiple paths of change in species abundances and ecosystem function (Zedler 2000). As ecosystems move through multiple states of degr adation (Hobbs and No rton 1996), there are disturbance thresholds that may be crossed, ma king reversal much mo re difficult, if not impossible (Whisenant 1999, Lindig-Cisneros et al. 2003). Recent studies have shown that feedbacks can develop between novel communities and degraded environmental conditions that result in resilience to re storative change, even a fter key structuring
13 processes (fire, nutrients, hydrology) hav e been restored (Bakker and Berendse 1999, Zedler 2000, Suding et al. 2004). For exampl e, fires may be ineffective at restoring prairie grass communities if invading woody s pecies impact burn efficacy, resulting in a novel composition (Anderson et al. 2000). Si milarly, overgrazing in semi-arid systems and subsequent shrub encroachment can lead to changes in soil characteristics and water availability that cannot be reversed by simply easing grazing pressure (van de Koppel et al. 1997). To further complicate matters, there are ma ny cases where historical structuring processes may no longer be restorable; natur al disturbances like fires and floods are often suppressed or eliminated entirely from the system. Prescribed burns are increasingly difficult to conduct near urban areas and historical flood levels in urbanized watersheds cannot be achieved without property losses. Essentially, the historical range of environmental variability in many ec osystems no longer exists, and restoring these key structuring processes may no long er be an option (Seastedt et al. 2008). Lakes and wetlands near urban areas are pr ime examples of such ecosystems, where a permanent loss of historical water le vel fluctuations and their crucial role in maintaining function and pattern is often ac companied by invasive or exotic species introductions and nutrient pollution (Havens et al. 1996). How can restoration succeed under such novel conditions? One appr oach is to find surrogates for lost processes; Infrequent, catastrophic fires in forests have been replaced by clear-cut logging (Hunter 1993) and spring-grazing by ca ttle has replaced frequent fires in prairie grasslands (Seastedt et al. 2008). In lakes and wetlands, managers have long used drawdowns in place of natural droughts, but, like prescribed burning, this becomes
14 increasingly difficult in large systems or wher e recreational activities are interrupted. Additionally, new communities establishe d under higher nutrient loads and stabilized water levels can reduce the efficacy of infrequent drawdowns (Moyer et al. 1989), maintaining dominance over more desirable communities even after drawdowns. In some cases, heavy machinery has been used to speed up organic sediment and biomass removal during drawdowns in an atte mpt to restore desired plant communities and substrate qualities to degraded systems (Moye r et al. 1995, Hoyer et al. 2008). In warm-water, shallow lake systems, dec ades of excessive vegetation growth under eutrophic and stabilized water conditions c an lead to thick, organic substrates and the establishment of highl y competitive plant monocultures. Dense vegetative growth can impede navigation, alter fish comm unities (Killgore et al. 1989) and foraging efficiency (Diehl 1992), and combined with exotic species introductions can have dramatic impacts on shallow, warm-water la kes (Hoyer and Canfield 1997). As with many aquatic ecosystems, the disturbance r egime (flood/drought cycle), species pool, and geochemical conditions of these system s are or will soon be outside of their historical ranges. To combat these effect s, an aggressive restoration approach has been implemented in several su ch systems across Florida. This technique, hereafter referred to as mu ck removal, involves exposing parts of the littoral zone with drawdowns, remo ving accumulated organic material, and regulating initial plant establishment with se lective herbicide treat ments. The goal of these projects is ultimately to improve fish spawning habitat, recreational access, and overall water quality by removing dense li ttoral vegetation and associated organic substrates, and reducing the abundance of rapid litter-producing species like cattail
15 ( Typha spp.) and pickerelweed ( Pontederia cordata ) in the near-shore littoral zones (Hoyer et al. 2008). While several studies of this techni que have focused on sport fish response (Moyer et al. 1995, Allen et al. 2003), littl e information exists about actual habitat changes following these restoration activities (but see Moyer et al. 1987, Tugend and Allen 2004). These types of extrem e management approaches represent the challenges facing restoration ecology in t he future, specifically where historical structuring processes cannot be restored. Many inland wa terbodies are or will soon be facing issues like those of many Florida lakes, and it serves as an excellent example of intensifying management efforts to replac e natural processes in aquatic system restoration. This paper will focus on the largest application of this muck removal application to date, which took place on a central Florida lake, Lake Tohopekaliga. Lake Tohopekaliga Lake Tohopekaliga (hereafter referred to as Lake Toho) is one of several large lakes located in the upper Kissimmee River bas in, collectively draining thousands of square kilometers into the Kissimmee River and ultimately Lake Okeechobee (Figure 11). Lake Toho and an adjacent sister lake, East Lake Toho, are the northernmost lakes in the basin, lying between t he Orlando and Mount Dora Ridges in the Osceola Plain. This plain consists mainly of poorly drai ned, clayey sediments with poor groundwater recharge, having over 73 lakes at least 3.2 ha in size (HDR Engineering 1989). Most of the lakes in this region were formed from solu tion activities and are precipitation driven. Lake Toho is the largest lake in the O sceola Plain, covering an area of 8,176 ha with an average depth of 2.1 m at maximum pool (16.75 m NGVD) (HDR Engineering 1989, Remetrix LLC 2000). The immediate watershed is 340 km2, though an additional
16 686 km2 of East Lake Toho watershed ultimately drains into Lake Toho through canal C-31 (HDR Engineering 1989). Nearly half of these 1334 km2 are drained primarily by two main stream systems: Shingle Creek, lo cated north of Lake Toho and flowing directly into the northwest side of the lake; and Boggy Creek, northeast of Lake Toho and flowing into East Lake Toho. Depen ding on precipitation and the operation of control structures on C-31 (drainage canal fr om East Lake Toho to Lake Toho) either Shingle Creek or the discharge from East Lake Toho can account for as much as 50% of the inflow to Lake Toho (Fan and Lin 1984, HDR Engineering 1989). Lake Okeechobee Orange County Osceola County Kissimmee River Florida East Lake Toho Lake Tohopekaliga N E W S 5km Lake Okeechobee Orange County Osceola County Kissimmee River Florida East Lake Toho Lake Tohopekaliga N E W S 5km 5km Figure 1-1. Location of Lake Tohopekaliga and East Lake Toho in relation to Lake Okeechobee and the Kissimmee River.
17 Brief History Historically, much of the watershed in the upper Kissimmee River basin was dominated by wetlands, with lakes bordered and interconnected by large wet prairie sloughs, including the connection of Lake Toho and East Lake Toho by Fennel and Cross Prairies (HDR Engine ering 1989). This network of waterbodies flowed south primarily through the Kissimmee River, virtually connecting waters of interior central Florida to Lake Okeechobee. As early as the 1850s, pioneers began to m odify the hydrology of the system and by 1884 a navigable waterway was opened from Kissimmee all the way to Fort Myers (HDR Engineering 1989). After the Flori da Legislature passed the General Drainage Act in 1913 (Chap. 298, FS), a reported 108 km (67 mi) of canals were dug throughout the Shingle and Boggy Creek Basins (Blackman 1973). Catastrophic hurricanes in the 1940s sparked several flood control projects with major changes occurring in the upper Kissimmee River basin by 1957. These projects were designed to construct levees and control structures on the south ends of t he larger lakes, to improve channels to downstream lakes, and for regulation of upper lake levels within a 0.6.2 m range (HDR Engineering 1989, U.S. Army Corps of Engineers 1956). Water control structures and canals regulating flows to and from Lak e Toho were completed in 1964 (Blake 1980), marking the end of natural water level fluctuations. This resulted in a stage reduction from at least 3.2 m to a maximum of 1.1 m (Wegener et al. 1973). Figure 1-2 shows the sharp contrast bet ween the dynamic, astatic condition prior to impoundment in 1964 and the stabilization that has occurred since.
18 14.50 15.00 15.50 16.00 16.50 17.00 17.50 18.00 6/1/405/30/19505/27/19605/25/19705/22/19805/20/19905/17/2000 Regulation Begins Daily Elevation (meters NGVD)1945195519651975198519952005 16.0 17.0 18.0 15.0 Figure 1-2. Daily mean water elevations in meters (NGVD) from January 1942 until Jun 2008. The dashed gray vertical line represents the approximate time of impoundment in 1964 while the dashed circ les indicate managed drawdowns. The natural drought in 1962 and flood in 1960 were the lowest and highest on record at that point. Sewage treatment plants began pumping ef fluent into the Shingle and Boggy Creek basins as early as the 1940s, and by 1986 an estimated 113 million liters per day (30 million gallons) were being discharged in to these systems (Wegener et al. 1973). Though water quality problems were recognized and attributed to these plants in 1969 (Wegener 1969), discharges were not completely eliminated until 1988. By that point nutrient loading and water level stabilizati on had noticeably affected littoral habitats, water qualities, and fish populations (Moyer et al. 1989). Mean lake phosphorous (total)
19 levels dropped 85% from 1980 to the mid 1990 s (0.82.11 mg/L), while total nitrogen dropped 50% (2.69.30 mg/L) (Williams 2001). Previous Studies In 1969, the Florida Fish and Wildlif e Conservation Commission (FFWCC) recommended that all effluent discharges into Lake Toho be stopped and that a managed drawdown be performed in hopes of sparking seed germination and recolonization of desired species (Wegener 1969). The first managed drawdown of the lake took place in 1971, lowering the wate r from a high pool stage of 16.75 m to 14.65 m (55 ft) NGVD (National Geodetic Vertic al Datum). The lake was held there for nearly six months and drought conditions fu rther extended the refilling to high pool stage until March of 1973. During this period the FFWCC conducted studies on fish, invertebrates, vegetation, soils, algae, and water chemistry (Wegener et al. 1973). Vegetation studies consisted of fixed sa mpling along line transects established perpendicular to the shore, ranging from above high pool stage to the lakeward extent of emergent vegetation. Frequencies of occu rrence of species were recorded based on a form of line intercept method using a fivepointed rake (Sincock et al. 1957). At that time the only vegetation considered a nuisance was water hyacinth ( Eichhornia crassipes ) and the overall expansion of littoral co mmunities into the lake by 16% was hailed as a success (Wegener and Williams 1974). Another drawdown was performed in 1979 based on the successes of the previous effort. Sport-fish populations increased to a maximum by 1982 and then gradually declined to the lowest level since 1972 (Moyer et al. 1989). Based on these data it was assumed the habitat had degrad ed substantially and would no longer support maximum fish densities. No vegetation studies were conducted.
20 In 1987, the discharge of effluent to the lake was almost eliminated and another drawdown was performed. Contrary to the ot hers, which were implemented to increase the density and area of the littora l zone in general, the purpos e of this project was to eliminate dense, monocultura l stands of vegetation ( Polygonum spp. and Pontederia cordata ) that had formed an organic barrier from accumulated organic matter; isolating many shallow areas of littoral zone to the point of blocking acce ss of sport fish to important spawning grounds (Moyer et al. 1989). The goal of the 1987 dry down was to reestablish native grasses in place of these dense, monocultural stands of unwanted vegeta tion. This marked the first mechanical muck-removal project, scrapi ng approximately 172,000 m3 of muck and vegetation from various sections of shorelines. After ju st two years, however, line transect studies established in 1986 showed an almost comple te rebound of the vegetation targeted for removal ( Pontederia cordata ), though several grass specie s increased in frequency as well (Moyer et al. 1989). A natural drought in 1991 gave lake m anagers another opportunity to remove some of the unwanted vegetation and two re moval experiments were performed, one involving mowing the vegetation to a maximu m height of 15 cm and the other, uprooting and removing it. It was found that Pontederia rebounded in both tr eatments, though at a slower rate after uprooting. Herbici de applications were also made in hopes of minimizing the regrowth of Pontederia but were only effective at slowing regrowth. Study Objectives and Methodology Muck removal projects can provide im portant information to managers and restoration practitioners beyond the specific effects of individual projects. For one, they are prime examples of aquatic gardening, or using increa sed external inputs/efforts to
21 maintain or establish a desired pattern in a system, rather t han restoring the key structuring processes that may have produced that pattern originally (Mitsch and Wilson 1996). Large-scale applications also prov ide an opportunity to implement adaptive management strategies, comparing preand pos t-restoration communities and testing the efficacy of new management activities before applying to other systems. Additionally, they provide an opportunity to study how aggressive management techniques perform at large scales; i.e. how well the results conform to predicted effects, and what might impact such outcome s. These are all important issues in restoration applications, and t he following chapters of this paper will attempt to address them. These issues can be broken down into three primary questions, which are further discussed throughout the document: Can aggressive, intensive management e fforts re-establish desirable littoral vegetation patterns without t he structuring processes (hydrology, nutrients) that maintained those patterns historically? By comparing preand post-restoration littoral vegetation communities, what are the effects of this technique and how mi ght these results be incorporated into future applications? How well did vegetation responses to a large-scale restoration project follow managers predictions? Did environmental disturbances immediately following muck removal lead to unexpected result s? What factors might need to be addressed before implementi ng expensive, intense restoration efforts? Several study sites and sample designs were implemented to address these questions, and an outline of the differences is included below to aid interpretation. Pattern without Process The ability to re-establish vegetation patterns without historical structuring processes (Chapter 2) was studied on a variet y of shorelines throughout the lake. Sampling occurred within the ent ire emergent littoral zone at these sites, both before
22 and after the project was initiated. The pr imary difference from the sites addressed in Chapter 3 is that sampling was conducted in both shallow and deeper sections of the emergent littoral zone, not all of which we re mechanically scraped. Substrate quality and water levels limited the extent to wh ich heavy machinery could drive onto the exposed lake bottom, which essentially re stricted muck removal to an elevation of roughly 135 cm in water depth at maximum lake stage. Emergent vegetation extended tens or even hundreds of meters beyond thos e depths, to roughly 200 cm in depth at full pool. Thus, some of the samples were lo cated in shallower, scraped sections of shoreline, and others were located in deeper, unscraped sections of shoreline. These scraped/unscraped designations ma y be confused with treatment and control plots that are addressed in Chapter 3. T he important distinction is t hat the lake-wide study sites were all mechanically scraped, unless they occurred in areas too deep for muck removal. Thus, there are no control sample s in these sites, only scraped (shallower) and unscraped (deeper). Mechanical Muck Removal Effects The preand post-restoration compar ison of vegetation communities, or restoration effects, were der ived from experimental study sites located along specific sections of shoreline that were on simila r slopes and dominated by similar communities. Sampling within these sites was restricted to only those depths that mechanical removal could occur, and did not include the deepe r-water, unscraped sections of emergent littoral zone that are included in Chapter 2. Throughout Chapters 3 and 4, these st udy areas will be referred to as experimental sites, which contained both control and treatment pl ots. These study areas had sections of shorelin e specifically deli neated as control plots, which were not
23 mechanically scraped at any depth. The contro l and treatment plots in the experimental sites are not to be confused with the lakewide sites that also contain scraped and unscraped samples, which refer to differences in water depth rather than experimentally applied treatments. It should al so be noted that in the experim ental sites, control plots were not necessarily devoid of all treatment, as they were still subjected to the managed drawdown that occurred throughout the lake. Effects of Environmen tal Disturbances on Muck Removal Project Chapter 4 draws on results from both the lake-wide and experimental study sites discussed above. There are also references to preand post-water-level regulation periods, as well as pre-and post-restoration. Water-level regulation occurred in 1964 and the restoration projec t took place in 2004. Both pr e-and post-restoration vegetation communities are compared to pr e-regulation (historical) co mmunities, and care should be taken as to which event the preand postdesignations are referring to. Throughout this paper, vegetation composit ions will typically be classified as communities, which in this case are simply a collection of species found at a specific place and time. Some studies have used the term in an abstract sense, referring to communities as a working mechanism or org anism (e.g. Watt 1947), but throughout this paper it will be used in a concrete sense, st rictly for classification purposes (e.g. McCune and Grace 2002). Thes e groupings will represent spec ific abundances of one or more dominant species relative to one another along measured environmental gradients.
24 CHAPTER 2 RESTORING LITTORAL VEGETATION PA TTERNS WITHOUT KEY STRUCTURING PROCESSES Introduction Restoration ecology has evolved from simple, succession-based approaches (van der Valk 1998) to addressing multiple cons traints and degradation thresholds (Suding et al. 2004) as more complex systems and projec ts are undertaken. Regardless of the system or subject of restorati on, the primary task is usually to re-establish historical structuring processes or key variables t hat have been lost and re sulted in degradation (Mitsch 1998, Hunt et al. 1999) This approach of restori ng the structuring processes necessary to re-establish a desired structure or species mix (pattern), relies on selforganization and self-sustenance of communiti es, minimizing external influence or energy inputs (Parker 1997). The frequency and intensity of fires, floods, and droughts (disturbance regimes), for example, are often t he focus of restoration in forest (Cissel et al. 1999), prairie (Johnson and Matchett 2001) and wetland ecosystems (de Angelis 1998) because of their importance as key structuring processes. The natural or historical range of vari ability of these processes have long been used as guides to maintain diversity, restore ecosystems, and influence future management actions (Landres et al. 1999). Ho wever, there are increasing numbers of systems where key structuring processes, or the historical range of environmental variability no longer exists, and restoring thos e conditions may not be economically or socially feasible (Seastedt et al. 2008). Fo r example, large-scale intense burn regimes that may have occurred historically cannot be safely implemented by managers, and even small-scale, low-intensity burns may be unacceptable near urban areas (Stephens and Ruth 2005). Many aquatic systems have altered flood/drought cycles that are
25 largely determined by water use or flood control needs (Hill et al. 1998, Havens and Gawlik 2005), and restoring historical cond itions would jeopardi ze surrounding urban developments. When key structuring processes are los t, external inputs (management efforts) must increase to maintain or restore desir ed patterns, which can ultimately resemble intense gardening or land scape architecture more than ecosystem restoration (Mitsch and Wilson 1996). This can be a gradual occurrence; as species pools or disturbance regimes slowly depart from hi storical conditions, management efforts increasingly focus on removing unwanted species or the effects of those species (higher fuel loads, increased peat deposition). Event ually these methods can constitute the majority of management effort. This removal-based approach has been crit icized as not necessarily capable of restoring any historical state, or even establishing a new, des irable state (Seastedt et al. 2008). Additionally, there are a growing number of exampl es where under prolonged degraded conditions, biotic and abiotic interact ions create new feedbacks that result in resilience to restorative change, even if lost structuring processes could be restored (Baker and Berdense 1999, Zedler 2000, Sudi ng et al. 2004). Changing environmental conditions (climate, disturbance regime s, etc.) and new combinations of native/introduced species have led to the deve lopment of novel ecosystems (Suding et al. 2004, Hobbs et al. 2006) and management st rategies will have to be re-examined in many cases (Seastedt et al. 2008). While most examples of novel ecosystems in the literature are terrestrial, lakes and wetlands could easily be at the forefront of this issue. Watershed development and
26 land use has long affected the hydrologic sc hedules, nutrient loads, and species pools of virtually every aquatic system (e.g Davis and Ogden 1997), and managers have relied on removal-based approaches for decades to maintain desired patterns under altered conditions (e.g. exotic or invasive species removal). Current environmental problems resulting from past issues (regional expansion of exotics, unchecked nutrient loading) can consume the time and budgets of managers (Seastedt et. al 2008), so that new approaches and methods are slow to develop. Most of our major water bodies are or will soon be permanently outside of their natural range of environmental variability. Hydrologic schedules will likely be increasingly determined by growing wate r supply and flood control demands, while global climate change may affect the frequency, intensity, and duration of flooding/drying events (Abrahams 2008). These new challenges could render traditional management approaches unor even counter-pr oductive, and focus may have to shift towards finding desirable or acceptable communi ties that provide biotic structure and ecosystem services under new condi tions (Seastedt et al. 2008). This paper focuses on a lake restoration pr oject in central Florida, USA, as an example of the challenges facing many aquati c restoration projects in novel systems, and of the need to develop new management approaches. Classic removal-based management efforts have not been able to pr event degradation from various water, nutrient, and species pool changes in this large, shallow, sub-tropical lake and increasingly expensive and intensive met hods of habitat manipulation have been developed as a result (Moyer et al. 1995).
27 This paper focuses on a newer, intensive approach to lake restoration, where bulldozers, herbicides, and drawdowns are used to maintain desirable, native vegetation communities in a system now outsi de the range of natural variability. This technique, dubbed "muck removal", was developed as a method of removing accumulated organic material from the littoral zone, deposited after decades of excessive vegetation growth under eut rophic and stabilized water conditions. There is a long history of managers combat ing dense growth of aquatic vegetation (Holm et al. 1969), as it can alter fish comm unities (Killgore et al 1989), lower oxygen levels, and impede navigation an d flood control (Little 1968). Faced with increased difficulty in implementing large-scale dr y-downs, coupled with the rising costs, inefficacy, and unpopularity of broad-scale herbicide applications, managers developed a single, intensive disturbance effort to restore desirable communities. This approach involves drying parts of the littoral zone, removing accumulated organic material and unwanted, resilient species from the shorelines, and r egulating initial vegetation succession with herbicides. The goal is to create a sandy bottom habitat with sparse emergent/submergent vegetati on that supports important sportfish spawning and improves boater access (Moyer et al. 1995). T hese conditions are considered historical targets for muck remova l restoration and are typical of systems with lower productivity or higher levels of disturbance than what is seen in this degraded system (see Grime 1979). This study represents far more than one states unique approach to combat decades of habitat degradation; it serves as an example of how systems continually change when important structur ing processes are lost, and how management agencies
28 often rely on increasingly expensive, remova l-based techniques despite criticisms of these approaches for nearly a decade (Hol ling 2001). This chapter addresses one important question facing re storation ecologists and pos es another; can historical vegetation structure or pattern be restored under novel conditions? If so, at what cost should we maintain desired patterns when current processes no longer support them? To answer the former question I compared v egetation communities after muck-removal restoration (scraping) to early vegetation patterns before water levels were regulated in the 1960s. Specifically, I compared domin ant species from the 1950s to those established in this study four years after rest oration. It is my hope these results can be used to address the latter question, and perhaps reemphasize the need to shift our restoration targets and find new methods to manage ecosystems under increasingly novel conditions. Methods and Analyses Lake Tohopekaliga (hereafter referred to as Lake Toho) is one of several large lakes in the Upper Kissimmee River Basin in c entral Florida, USA. Each of these lakes is successively connected and jointly regulated to collectively drain thousands of square kilometers into the Kissimmee River and ultimately Lake Okeechobee (Ch. 1, Figure 11). Toho is located near the top of this chain of lakes, covering 8,176 ha with an average depth of 2.1 m (at maximum stage regulation of 16.75 m NGVD), with a watershed of roughly 334 km2. Its volume has a large impact on water bodies downstream when drawdowns ar e implemented, especially si nce all lakes farther down the chain must be lowered to gravitati onally drain Lake Toho (HDR Engineering 1989, Remetrix LLC 2000).
29 Water control structures were install ed on the Kissimmee Chain and Lake Toho in 1964 (Blake 1980), reducing the variability of la ke stages by approximately 1.5 m, to an annual range of 1.1 m (Ch. 1, Fi gure 1-2). Sewage effluent from up to three treatment facilities was discharged into the lake from roughly 1940 to 1988, reaching a maximum of 113 million liters per day in 1986 (Wegener et al. 1973). Decreases in water quality and fisheries production in the late 1960's were attributed to stage regulation and nutrient loading (Wegener 1969), beginning a se ries of drawdowns in 1971, 1979, 1987, and 2004 as a means of improving fish habitat. Lake-stage data were co llected from a South Florida Water Management District gauge located on the southern end of the lake, recording average daily water levels since January 1942. Indicators of Hy drologic Alterations (IHA) software was used to analyze pre (1942) and post waterlevel regulation (1965) periods, defined by the completion of water control structures in 1964. Both periods were analyzed for 30, 60, and 90 day maximum and minimum water levels, average monthly levels, and quartiles. Two 10 year periods before and a fter regulation, 1944 and 1992, were selected for further com parison between preand post-regulation periods. This was done to exclude some of the more extreme events in either period, including historical highs and lows experienc ed in the early 1960s, as well as managed drawdowns in the latter period. Water leve l structures were not completed on the Kissimmee Chain of lakes until 1964, but it seems reasonable that substantial changes were occurring in the watershed up to severa l years prior (beginning of construction), possibly contributing to the historical highs (1960) and lows (1962) recorded just prior to completion. Rather than use these events as representative of typical pre-regulation
30 fluctuations, the 10 year period characteriz ed by more modest events and that occurred just prior to early vegetation studies (1956) was chosen. These two periods allowed comparison of relatively stable events, ra ther than just looking at differences in maximum and minimum stage levels, which woul d also be affected by rare climatic events. Water years were defined as June through May of the following year, corresponding with the beginning of the wet season and end of the dry season, respectively. The largest muck removal project to dat e was implemented on Lake Toho in 2004, when water levels were dropped from 16.8 m to 14.9 m from November 2003 until June 2004. Approximat ely 6.5 x 10^6 m3 of muck and vegetation were removed from over 80% of the shoreline and deposit ed at several upland sites, as well as 29 locations within the lake (FFWCC 2004). Water levels returned to normal by August 2004. Sample Locations Lake Toho has a variable littoral zone in terms of slopes, wave energy, and shoreline activities, which may result in diffe ring plant communities. In order to capture this variability and provide lake-wide inference for treatment effects, five monitoring sites were selected from the lessdeveloped, southern two-thirds of the lake that were uninterrupted by outflows, jetties, or other not able features (Figure 21). The goal of site selection was to represent the various habi tat types and geomorphic conditions of the lake without complicating restoration res ponses with shoreline development or unusual features (creek outflows) within the plots. Sites 1, 3 and 4 were located on broad, gent ly sloping areas of shoreline, while Sites 2 and 5 were located on steeper, highe r energy shorelines. Cattle ranching and grazing was the primary land use in all but sites 4 and 5, which were bordered by
31 infrequent residential housing. These locati ons represented the majority of shoreline community types and primary land use, giving the best inference as to lake-wide littoral response to treatment. # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S N E W SSite 3 1 1 0 0 m e t e r s Site 4 Site 1 Site 2 Site 5 Site 3Sampling locations 0 6 ft depth classes # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S N E W SSite 3 1 1 0 0 m e t e r s Site 4 Site 1 Site 2 Site 5 Site 3 # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S # S N E W SSite 3 1 1 0 0 m e t e r s Site 4 Site 1 Site 2 Site 5 Site 3 Site 4 Site 1 Site 2 Site 5 Site 3Sampling locations 0 6 ft depth classes 0 6 ft depth classes Figure 2-1. Location of study sites thr oughout the lake. Each site encompassed the entire width of the emergent littoral zone, to appr oximately 2 m in depth. Sample locations were randomly loca ted within depth stratifications of approximately 35 cm each. Samples were collected from all water depths occupied by the emergent littoral zone in order to monitor lake-wide veget ation changes, including those areas too deep for mechanical removal but still subjected to drawdown. The boundaries of each site were determined by placing a 60-ha rectangl e on Digital Ortho Quarter Quads (DOQQ) with 1-m2 resolution (1999) and bathymetric (Rem etrix 2000) layers in ArcView GIS 3.2
32 software. The area of the rectangle remain ed constant but the shape was altered such that it encompassed the zone of 0 m in dept h (0 ft) (i.e., the sites on steep slopes were stretched along the shore while those on gentle slopes extended much farther into the lake). Sampling locations were stratified by five depth classes of approximately 35 cm (one foot) each (One = 30 65 cm, Two = 65-100 cm, Three = 100 135 cm, Four = 135 170 cm, and Five = 170 205 cm). The three shallowest classes (30 135 cm) were located within the areas slated for mec hanical muck removal, while the majority of deeper classes (135 205 cm) were never dr ied enough or did not contain vegetation communities targeted for restoration. Random coordinates were used to locate samples in the field with a GPS (Global Po sitioning System) on each collection. The shallowest areas sampled coincided with t he approximate shoreward extent of muck removal activities, which was roughly the 30 cm depth contour at maxi mum lake stage. Three locations per depth class were sampled at each site for a total of 15 samples per site, 75 total. Aboveground vege tation was clipped from 0.25 m2 circular quadrats at each location, sorted by species, stems counted, and biomass recorded after excess water was shaken off. Samples were collected each year during the summer (June) and winter (December) seasons from June 2002 June 2008, to account for any seasonal variations in response and to in crease the temporal intensity of sample events. This resulted in 13 repeated measures for each location. Lake-stage data were colle cted from a South Florida Water Management District gauge located on the southern end of the lake, recording average daily water levels
33 since January 1942. Average water depths from three points at each sampling location were referenced to lake stage data to approx imate shoreline elevations across sites. Historical vegetation data were refe renced from agency studies conducted in 1956, eight years prior to water regulatio n (Sincock et al. 1957) and in 1971, seven years after (Holcomb and Wegener 1971). These studies used intercept methods to calculate frequency of occurrence along transec ts perpendicular to shore, with several locations overlapping the sites in this study. While their methods vary considerably from the sample techniques used in this study, t hey represent the only estimates of dominant species and their distributions along shore line elevations prior to stage regulation. Analyses Pre/post restoration comparison Plant species density and biomass were summed across sub-samples within depth classes for each sample event, resulting in five samples per site and 25 per sampling occasion. While each depth cla ss represented a subsample within a study site, they were analyzed separately to determine water depth influences on vegetation response. The December 2002 and June 2003 sample events were used to represent pre-restoration conditions, and December 2007 and June 2008 samples were used as post-restoration representatives. These time s were chosen to capture littoral conditions immediately prior to rest oration and after vegetation communities re-colonized mechanically-scraped areas. Importance Values (IV) were computed for each species from the summed sub-samples by averaging the relative biomass and density of each species in those sub-amples, and converting to a percentage value IV = (Relative Biomass + Relative Density)/2 *100
34 Importance values were used to estimate species importance within a given depth class and site as they are not biased toward s large, few-stemmed species or small, numerous-stemmed species. This calculation also relativized the dataset, eliminating the need for transformations typically applied to density or biomass data that can vary by orders of magnitude between specie s and samples (McCune and Grace 2002). To reduce noise from rare species, onl y those with cumulative IVs composing 99% of the total were retained for this analysi s. Many studies el iminate species that occur in < 1% of samples, but I used 1% of t he total IV instead. This method is more representative of the actual abundance of a species for the same reason IVs are more representative of a specie s abundance than frequency. The resultant matrix consisted of 100 (50 pre and 50 post) samples and 21 species of the 37 encountered, and includ ed a summer and winter sample event for both the preand post-restorat ion periods. Although only 5 of these 100 samples were independent (sites = 5), eac h depth class (5) and sampling occasion (4) were analyzed separately in order to account for gradient responses and seasonal variation within study sites. All analyses for the pre/post comparisons were performed on this matrix and unless otherwise specified, were perfo rmed using PCORD software (McCune and Mefford 1999). Shannon-Weiner diversity indices were calc ulated from species biomasses prior to eliminating rare species. The index was converted to effective numbers of species (exponential of the diversity index), which is useful when comparing across groups due to the non-linear nature of diversity indices (Hill 1973, Peet 1974). The effective number
35 of species in this case is essentially a m easure of how many species were consistently abundant relative to other s pecies during that period. Groups with similar species compositions were identified usi ng a hierarchical, agglomerative cluster analysis with flexible beta (-0.25) linkage and Sorenson distance measures. These were chosen for their sp ace conserving properties, compatibilities with each other, and their advantages with no n-normal data (McCune and Grace 2002). Samples were grouped based on species IV's, using multiple species as a basis for deciding on the fusion of additional groups. An Indicator Species Analysis (ISA) det ermined the optimum number of clusters for further analysis and defined those clusters in terms of representative species. This analysis uses the proportional IV and frequency of a particular species in a particular cluster relative to its IV and frequency in all other clusters (Dufrene and Legendre 1997). The optimum number of clusters wa s determined by which level produced the most species with p-values < 0.05 in the indicator species analysis (McCune and Grace 2002). Species with low p-values (< 0.05) a nd high indicator values (> 40) were used as community descriptors (clust er labels) in future analyses. A Classification and Regression Tree (CART) model (S-Plus Tree Library, Death 2002) classified samples into the groups id entified by the cluster analysis using the measured environmental variables alone. We used the Gini index to measure impurity (Breiman et al. 1984), cross-valid ation to select the best tr ee size (Vaysieres et al. 2000), and 1-Cross Validated Error as a more conservative estimate of variation explained by the model (De'ath 2002). CART models were built using the combined matrix of preand post-restoration comm unities in order to determine how muck-
36 removal effects varied temporally, s patially, and with water depth. Several combinations of the variables water depth (at full pool) and hydroperiod were used as continuous predictors, while site, scraping (f or the post-restorat ion period), dominant land use (cattle grazing, yes/no), and whet her the sample occurred on a floating mat were used as categorical variables. A Multivariate Regression Tree (MRT) was bu ilt to identify communities based on species IVs and where they occurred along environmental gradients, and to compare the results with those from the cluster analysis and CART model. This method uses the sum of squared Euclidian distances about t he multivariate mean of samples as an impurity measure of each node, and each split is made to maximize the sum of squares between nodes and to minimize it within no des (Death 2002). Each leaf is then characterized by the multivar iate mean of its samples, the number of samples within that leaf, and their defining environmental va riables. The percent of variation explained by the tree is reported as 1 Relative Error, or more strictly 1 Cross Validated Error. In short, this technique partitions the sa mples into communities using both species IVs as well as the associated environment al variables, and provides the threshold values for each partitioning variable. The resu ltant communities are defined not just by species compositions (as in the cluste r analysis) but where they occurred on the environmental gradients as well, provid ing a different approach to community delineation than the combined cluster and CA RT model methods. This method was used as a confirmatory procedure, to veri fy groupings and distri butional thresholds identified in the previous analyses.
37 Nonmetric Multidimensional Scaling (NMS ) ordinations were used to graphically display community shifts from preto pos t-restoration periods using IVs (McCune and Grace 2002). Pearson correlations were calculated for species biomass and water depths for each of the ordination axes. Vect ors were used to track species composition changes over time within sample units (sample position in species space), and approximate boundaries were drawn onto ordinat ion results to represent changes from one community type to another, based on the af ore-mentioned cluster analyses. Depth classes were graphically separated from a single ordination to ease interpretation. Temporal analyses In order to track change in species bioma ss over time, I combined sub-samples of biomass within each depth class, and then follo wed changes in species composition by site, for two periods prior to restoration and fo r the last two periods of the study. Each sampling period (pre = December 2002 and June 2003, post = December 2007 and June 2008) consisted of 25 samples, five depth cl asses at each site of five sites. These data were ordinated using NMS to graphically display changes in species abundances throughout the period of study. Results were graphically separated after the ordination by depth class for ease of interpretation. Results Hydrological Comparisons to Historical Record Following regulation in 1964, annual fluctuations were managed between 15.9 and 16.8 meters NGVD, with flood events reduc ed by approximately 1.0.5 m (Ch. 1, Figure 1-2). Average monthly water levels were similar when comparing preand postregulation periods, except for the early gr owing season (Figure 2-2). February and
38 March water levels were held higher and for longer duration after 1965, and dropped more quickly to low pool by the end of the dry season (June). The 10 year pre(1944) and pos t-regulation (1992) periods had similar mean, annual lake stages (16.45 and 16.40 m, respectively) and within-year variation (0.61.91 St Dev and 0.63.01 St Dev respectively) (Figure 2-3). However, inter-annual variations were considerably lowe r in the post-regulation period, with the standard deviation of annual means falling from 1. 09 prior to regulation, to 0.22 after. The largest reduction in between-year variat ion occurred for the months of September and October, where pre-regulation water levels for these months varied by 2 standard deviations. From 1992 to 2002, standard deviati ons for these fall months were 0.26 and 0.34, respectively.
39 17.0 16.25 16.0 16.75 16.5 15.75Lake Stage (Meters NGVD)JunJul Aug SepOct NovDecJanFeb Mar AprMay Figure 2-2. Monthly lake stage (Meters NGVD) for the pr e-regulation period 1942 (green), and post-regulation period 1965 (red). The 75th percentiles are shown for the pre-regulation period.
40 15.00 15.50 16.00 16.50 17.00 17.50 18.006/1/19446/1/19475/31/19505/30/19535/30/19935/29/19965/29/19995/28/2002 1953 1949 1944 1994 19982002 Pre-regulation 1944-1954 Post-regulation 1992-2002 15.5 16.0 17.5 16.5 18.0 17.0Daily Elevation (meters NGVD) Figure 2-3. Daily lake stage (Meters NGVD ) for a pre-regulati on (1944) and postregulation period (1992). Vegetation Response Total numbers of species encountered in t he two preand two post-restoration samples were 32 and 27, respectively, with frequency histograms showing similar distributions among samples between the two per iods (Figure 2-4). The majority of samples had 2 species in each quadrat, regardless of sample period. A histogram of scraped samples showing s pecies changes over time displayed a similar pattern, with the majo rity of quadrats having no change or a loss/gain of only one species (Figure 2-5). When deep er water, unscraped samples (still exposed to artificial
41 dry-down) were included in the histogram there was a slight left skew, but no strong pattern in overall species richness was evident throughout the study period. 0 5 10 15 20 25 123456789 Pre (32) Post (27)Number of SamplesNumber of Species by Sample Figure 2-4. Frequency histogr am displaying number of s pecies per quadrat for all samples before and after restoration. 0 1 2 3 4 5 6 -5-4-3-2-10123Number of samplesSpecies Change by Sample 0 5 10 15 20 25 -6-5-4-3-2-10123456 Number of samplesSpecies Change by Sample A B Figure 2-5. Frequency histogram of species gain or loss by quadrat after restoration. A) Only those samples that were me chanically scraped, roughly 30 cm in depth. B) All samples, including thos e too deep for scraping but still exposed during the drawdown.
42 Shannon-Weiner diversity indices showed an increase in the number of effective species in the scraped areas, fr om 5.3 to 6.2 after restorat ion. However, when including those samples too deep for mechanical removal, there was a decline in diversity from 7.9 to 5.5 effective species, highlighting a lo ss from those areas that were only subject to drawdown. The cluster analysis revealed six distinct groups based on the number of indicator species at each level of clustering, whic h resulted in approx imately 50% of the information remaining based on a scaled Wi shart's objective function (McCune and Grace 2002). Groups were labeled with specie s that had indicator scores of roughly 50 or greater in a given cluster (T able 2-1). These groups were 1) Pontederia cordata, 2) Typha spp., 3) Hydrilla verticillata 4) Paspalidium geminatum, 5) Panicum repens and Eleocharis spp., and 6) Vallisneria americana Table 2-1. Indicator values (0) of spec ies in the six groups identified by the cluster analysis. Those species with values of approximately 50 and greater were chosen as representatives of the corresponding group, or community. Species are coded as the first 3 letters of genus and first 2 of specific epithet ( Pontederia cordata = PONCO). P-value Species Group 1 2345 6 0.0002 PONCO 71 12000 0 0.0002 TYPSP 0 95 000 0 0.0002 HYDVE 04 50 250 9 0.0002 PASGE 0017 73 0 3 0.0002 PANRE 8000 81 0 0.0002 ELESP 1100 48 0 0.0002 VALAM 00002 92 Community correlations to measured environ mental variables were determined by CART models. The best model was pruned to nine leaves based on the 1 cross validated error method, and used water depth at maximum lake stage, muck removal (y/n), sample period (pre/post-restorati on), and whether the sample was on a grazed
43 shoreline (Figure 2-6). The CART model had a cross validated error of 0.36, and a misclassification rate of 17%, explaining 64% of the variation in the dataset. Water depth at maximum lake stage was the most im portant predictor va riable, followed by muck removal, sample time and cattle grazing (yes or no). Sample time was incorporated into this analysis as a surrogate for the artificial drydown that all samples experienced, regardless of whether or not their water depths allowed mechanical removal of soil and vegetation. Samples that changed community types following the restoration but did not actua lly get scraped were considered to exhibit dry-down effects. Each of the six communities identified in the Cluster and ISA analyses were classified in the CART model, with Typ ha and P. geminatum having the smallest number of samples (4) and Vallisneria having the most (22 & 4). While water depth explained more of the variat ion in the model, muck removal was responsible for the primary division of the data. Vallisneria was predicted to occur in scraped areas of shoreline at depths > 98 cm up to approximately 135 cm or the maximum scraped depths. P. repens and Eleocharis spp. dominated the shallowest scraped samples (approx. 30 cm). As expected, unscraped communities ( no muck removal) were much more variable, as they included pre-restoration samples as well as post-restoration samples too deep for mechanical scraping. Water depth differentiated the two, as the maximum depth scraped was approximately 135 cm. Seven of the leaves in the tree were classified as unscraped, with the ma jority of samples predicted as Pontederia (18) dominating 30 cm in depth.
44 Not Scraped Depth < 109 cmPreDepth < 98 cm Scraped> 127 cm > 160 cm < 166 cm >184cm PONCO (18) TYPSP (4) HYDVE (14) VALAM (4) HYDVE (14) PASGE HYDVE (10) PASGE (4) PANRE ELESP (10) VALAM (22) Post PONCO TYPSP HYDVE PASGE PANRE_ELESP VALAM Figure 2-6. CART model of pre-post restor ation periods (9 groups, CV error = 0.36, Misclass rate = 17%). Bargraphs under leaves represent proportion of samples in that leaf classified as the corresponding community. Numbers in parentheses under the leaves indicate how many samples are in each leaf. Rank of variable importance was water depth, muck removal, sample period, and grazing. Prior to restoration, the Hydrilla and the P. geminatum communities were predicted to occur intermittently at > 127 cm of water depth, either as individual communities or grouped as one. Following the restoration, Hydrilla distributions were somewhat reduced, with Vallisneria replacing that community in 127 cm of water, regardless of whether or not samples were scraped. Typha was classified in just four unscraped samples, at 109 cm in depth, all of which were later scraped. Essentially, the 18 samples classified as Pontederia 4 as Typha and 14 classified as Hydrilla (36 total), all
45 became Vallisneria (22 & 4) or P. repens and Eleocharis (10) after restoration, depending on water depth. These same data were used to track changes in species compositions of individual sample units over time using an NMS ordination (Figure 2-7). The final solution was three dimensional, accounting for 68% of the variance with 54% in the primary two axes. Total biomass per sample was correlated with Axis 2 (R2 = -0.41) and water depth with Axis 1 (R2 = 0.72), having a final stress of 16.8 and an instability of 0.002. These values lie within the ranges typically seen with ecological community datasets (McCune and Grace 2002). The ordination results were graphed in the two primary dimensions, and approximate community boundaries were drawn in to highlight shifts in community type over time. Water depth cla sses were pulled apart into separate graphs to ease visual interpretation of 25 different sample units movi ng through four time periods (Figure 2-8). Due to the Pearson correlations with ordinat ion axes, Axis 1 can be approximated with water depth (R2 = 0.72) and Axis 2 with decreasing total biomass (R2 = -0.41).
46 Pontederia Vallisneria Hydrilla P. geminatum P. repens Eleocharis TyphaAxis 1 -Water Depth (R2= 0.72) Axis 2 -Biomass (R2= -0.41) Figure 2-7. NMS ordination of two preand two post-restoration sample periods. Approximate community boundaries based on cluster groupings were outlined on the ordination for ease of interp retation. Pearson correlations (R2) with axes were: Water depth = 0.72 to Axis 1, Total bioma ss = -0.41 to Axis 2.
47 PONCO VALAM HYDVE PASGE PANRE_ELESP TYPSP Axis 2Axis 1All30 65 cm 170 205 cm 65 100 cm 100 135 cm135 170 cm PONCO VALAM HYDVE PASGE PANRE_ELESP TYPSP PONCO VALAM HYDVE PASGE PANRE_ELESP TYPSP Axis 2Axis 1All30 65 cm 170 205 cm 65 100 cm 100 135 cm135 170 cm Figure 2-8. NMS ordination of two preand two post-restoration sample periods. Approximate community boundaries based on cluster groupings were overlayed onto the ordinat ion, and arrows show movement of samples through species space over time. Lar ge movements indicate large changes in species compositions, and vice versa. Depth classes were pulled from the same ordination to ease interpre tation. Pearson correlations (R2) with axes were: Water depth = 0.72 to Axis 1, To tal biomass = -0.41 to Axis 2. Similar to the CART model, the s hallowest depth class went from a Pontederia community to P. repens following restoration, while the second depth class moved from either Pontederia or Typha to Vallisneria This depth class saw the largest change in species composition, as evidenced by t he amount of distance between the preand post-restoration samples. The 3rd depth class contained several communi ty types prior to the muck removal, including Pontederia Typha and Hydrilla ; all but the Hydrilla community changed to
48 Vallisneria following restoration. The fourth depth class was beyond muck-removal depths, but also shifted primarily to Vallisneria from either P. geminatum or Hydrilla after restoration. The deepest samples remained largely unchanged, grouped as either Hydrilla or P. geminatum communities. The MRT had a higher error rate than the CART model (CV error = 0.55) explaining only 45% of the variation in the dat aset (Figure 2-9). Including site location and sample season improved the fit slightly, but I was more interest ed in isolating how mechanical scraping and dry-down varied by water depth, and wanted to simplify the tree (minimize number of leaves) in order to ea se interpretation. Si te differences that were delineated in alternative models were not attributed to obvious anomalies (grazing pressure, shoreline slope, dominant wind dire ctions, etc.) and so were considered as local variations and not included as a variable in the final model. Even without accounting for more than 50% of the variation in species abundances, the MRT still showed very simila r results to the CART model and cluster analysis. Most of the dominant species in t he final nine groups were listed as indicator species, and their positions along the gradients were very similar to the CART model. There were essentially seven groups identifi ed in the nine leaves of the tree, with the primary differences between the MRT and CART being the absence of Typha the addition of Nuphar advena and the splitting of the Pontederia group to include a shallower Luziola fluitans community.
49 Not Scraped > 98cm < 123cm < 160cm <166cm Pre Post > 65cm LUZFL PANRE PONCO ALTPH TYPSP HYDVE NUPAD PASGE ELESP VALAM (10) (22) (12)(12) (4) (16) (4) (6)(14) LUZFLVALAM HYDVE PASGE HYDVE VALAM LUZFLPONCO HYDVE Scraped Figure 2-9. MRT model of pre-post restor ation periods (9 groups, CV error = 0.55). Bargraphs under the leaves represent t he proportion of each species in that group. Numbers in parentheses under the leaves indicate how many samples are in each leaf. Rank of variable importance was water depth, sample period, muck removal, and grazing. Only the most common of the 21 species are shown. The primary division was attributed to muck removal, with the scraped samples again dominated by P. repens and Eleocharis spp. at < 98 cm in depth, while everything in 98 cm was dominated by Vallisneria There were also similar groups among the unscraped samples, with the next division occurring at 123 cm in the MRT, as compared to 109 cm in the CART model. The unscraped samples at < 123 cm were dominated by either Pontederia or Luziola
50 the same as the CART model, but were fu rther divided at 65 cm in depth with the shallowest community classified as Luziola The deep water communities were also qu ite similar to the CART model, with Hydrilla P. geminatum and Vallisneria all dominating at varyi ng depths > 123 cm. The MRT again identified temporal shifts from Hydrilla to Vallisneria even in the unscraped samples, but also revealed a loss of Nuphar over time as well. Samples > 160 cm in depth were again dominated by either P. geminatum or Hydrilla as in the CART model. Biomass plots showed similar pattern s in dominant species change after restoration, and highlight ed the dramatic loss of floating leaf aquatics and Typha that were subtle in the analyses based on importance values (Figure 2-10) These communities were relatively infrequent in st udy samples, but both lost 100% of their sampled biomasses after restoration. Pontederia the most common species and primary target of the mechani cal removal project, lost 88%. Similar to the importance value analyses, Vallisneria had by far the largest increase in biomass among the species (from 0 to 14.6 kg), but Hydrilla and P. repens both exotics, increased biomasses 137% and 58%, respectively. P. geminatum was very prevalent in the deepest samples and as evident in the ordination tracks, showed very little change in biomass between time periods.
51 -40 -30 -20 -10 0 10 20 TYPSP PONCO Pads PASGE PANRE HYDVE VALAM -100% -2% -88% -100% 137%58%N/A Figure 2-10. Changes in biomass of commo n species (in kilograms wet weight), including Nuphar advena and Nymphaea odorata grouped as lily pads. Biomass gain or loss is also express ed as a percentage of pre-restoration biomass, unless a species was not recorded previously ( Vallisneria ). Discussion Historical Comparisons The primary goal of this rest oration project was to replace robust communities that became established after years of elev ated nutrient levels and stabilized water schedules, with communities more typical of the dynamic, lower-nutrient system that occurred historically (sparse macrophyte cove rage and sandy substrates). Ideally, this would be accomplished by re-establishing the range of environm ental variation, including nutrient levels (Moss et al. 1996) hydrological regime (Middleton 1999), and species pools (Cole 1999). While point-sou rce nutrient reductions (James et al. 1994)
52 and selective species removals (Moyer et al. 1989) have been the focus of past management activities on the lake, annual water regulation schedules have been unchanged for over 40 years. As early as 1969, managers recognized the need to incorporate larger fluctuations into annual schedules (Wegener 1969) but, like many ecosystems today, this was hindered by fl ood control and water use demands in the changing watershed. Instead, management effort s focused on species removal, rather than addressing changes in processes that may have led to their establishment. For example, over a 10 yr period ju st prior to restor ation (1992), the shoreline elevation oc cupied by the dominant Pontederia community was flooded 78 99.9% of the period. The deepest edge of this community essentially never dried, and the shallowest edge was flooded by up to 83 cm of water during high water events. During a typical 10 yr period prior to r egulation (1944), however, this shoreline elevation was flooded only 66% of that period, with wate r receding 18 vertical cm below the deepest edge of the community du ring low water events, and flooding the shallowest edge by up to 1.7 meters of water during high water events. Besides a reduction in the magnitude of wa ter level variations, there was also a substantial difference in the timing of annual fl uctuations after water-level regulation. Historically, high water events consistently occurred in early winter (October and November), and shifted to early spring (F ebruary and March) after stabilization. Additionally, the early spring le vels are now higher than they were historically, and drop more suddenly to match annual lows in May and June. Water levels in the early growing season can have a big influence on species compositions (Weiher et al. 1996) limiting germination to more flood tolerant species, or
53 promoting vegetative growth from established perennials. Likewise, the timing (van der Valk 1981, Seabloom et al. 1998), frequency, and duration of flooding (Squires and van der Valk 1992) largely influences species compositions in wetland systems. The dramatic change in magnitude, seasonality, and frequency of flooding/drying events are likely the primary reason for a shift in dom inant vegetation type si nce stage regulation. Under the dynamic lake-levels of the 1950's, the shoreline elevations in this study were dominated by grassy species, namely P. repens, Paspalum spp., and Luziola (Sincock et al. 1957). While this period and habitat was often referenced as the target for shoreline restoration on this lake (sandy substrate, sparse emergent grasses), the dominant species prior to regulati on was an invasive exotic grass ( P. repens ) that was originally introduced for grazing cattle (T arver 1979). This same species has had dramatic impacts in other Florida lakes (Smi th et al. 2004) and is continually regulated with herbicides by the Bureau of Invasive Plant Management (Schardt 1994). During the 1950's, however, P. repens was not considered a nuisance on Lake Toho, and was described as an important spec ies that provided cover during higher water (Holcomb and Wegener 1971). This a good example of the di fficulty in defining the natur al or historical state of systems (Landres et al. 1999), as well as how re storation targets can often be biased or value-laden perceptions of what a system should resemble rather than what it was prior to human-induced changes (Zweig and Kitchens in press ). In this particular case, the restoration target appeared to resemble the habitat structure (sparse emergent grasses) that was established in the 1950s, but there is no record of a similar, native-dominated community having ever occurred on the lake.
54 Prior to restoration activities in 2004, a grassy community like t hat described in the 1950s was found in my study areas, but was restricted to narrow bands near the annual high water lines (< 58 cm) and was dominated by P. repens and Luziola The lower elevations that it occupied historically were dominated by more competitive emergent species, primarily Pontederia and the invasive exotic Alternanthera philoxeroides neither of which were identif ied in the 1950's (Sincock et al. 1957). Four years after restoration, there was a marked decrease in the Pontederia community but it was not replaced by the grassy communities that ex isted historically. Instead there was a dramatic expansion of s ubmersed aquatics, whose com position and abundance were unprecedented in this system. Vallisneria a native, submersed species, do minated scraped sections of shoreline four years after restoration, although it was extremely rare pr ior to restoration. While this plant had been recorded in the 1950s, it was at lower elevations (deeper water) and at very low frequencies (0%) in compar ison to this study (Sincock et al. 1957). By 1970 Vallisneria had moved shoreward nearly 60 vertical centimeters, expanding into elevations similar to where it was found at the end of this study, but its frequency of occurrence was still only around 3% (Holcomb and Wegener 1971). Hydrilla another dominant submersed species in this study, expanded in both the scraped and unscraped areas fo llowing restoration, and was never recorded on Lake Toho prior to 1972. Historically, Lake Toho was described as having low alkalinity and was not considered to support many submersed plants, and water-level stabilization was predicted to have little impact on their abundanc e (Sincock et al. 1957). Since that
55 period, nutrient pollution, wate r stabilization, and the introducti on of the Perfect Aquatic Weed Hydrilla (Langeland 1996), submersed species ha ve occupied large portions of the lake, and have hindered navigation and ev en flood control (BIPM 2004). This is another indication of the magnitude of change si nce water level regulations took place, and how dissimilar post-restoration communiti es remained from those in historical records. Management Implications The muck removal application conducted on Lake Toho was the largest and most intensive removal event to date, and was used as a substitute for the loss of historic flood/drought cycles that would have natur ally flushed/oxidized accumulated organic materials or floating mats during such event s. This unique approach was an attempt to reset decades of succession by removing the species, seedbanks, and soils associated with degraded shorelines. Initial colonizati on was regulated with selective herbicide applications, in an effort to force historical patterns on the sandy, exposed shorelines without the historical processes that maintained the system originally. Ecosystem restoration will increasingl y involve finding new methods or mechanisms to produce desired pattern where key drivers have been lost (Williams 2007). Global warming and human population gr owth is pushing many systems outside of their historical range of variability, and ma ny lost processes may not be restorable. The prevalence of novel ecosytems, or t he existence of new communities evolving under changing environmental conditions and al tered disturbance regimes (Suding et al. 2004), may require shifting restoration target s from focusing on historical states to finding new, potentially more beneficial ec osystem conditions (Chapin et al. 2006).
56 This study is a good example of the difficult y in applying traditional, removal-based management approaches with increasing efforts, attempting to maintain historical patterns under degraded conditions. After 30 years of gradual ly intensifying management techniques (Wegener and Williams 1974, Moyer et al. 1989), the largestscale muck removal project ever implemented still failed to restore historical littoral patterns on Lake Toho. The dense, monocultural Pontederia community that dominated for at least 20 years after hydrologic and nut rient changes in the system promoted its expansion, was replaced by yet another novel, submersed community that was previously only documented at very low fr equencies (Sincock et al. 1957, Holcomb and Wegener 1971, Moyer et al. 1989). While the objective of establishing the historical pattern of sparse, emergent grasses was not realized, the community that did establish along t he treated shorelines had similar desirable attributes as the target habitat. For example, organic soils were replaced with sandy substrates, the submer sed communities provide important habitat for fish (Barnett and Schneider 1974) and wa terfowl (McAtee 1939, Sculthorpe 1967), and the removal of floating mats and dense emergent vegetation allows for better recreational access. When system degradation cannot be revers ed at an acceptable cost, managers should strive for the biotic structure and ec osystem services that stakeholders desire, while promoting communities that are both feasible and resilient (Seastedt et al. 2008). This project has been successful at the fo rmer, but the resilience or longer-term establishment of a submersed community at these lake elevations is unknown. The vertical range of shoreline occupied by the submersed community will likely compress to
57 only those areas with 100% annual hydroperiods under normal regulations (see Ch. 4), and aggressive emergent species have already been re-established in the shallow regions of this study. A r egional drought could expose most of the shoreline occupied by the newly established submersed communi ty, allowing rapid invasion by emergent species like Pontederia or perhaps more likely, P. repens ; this species has previously displaced thousands of hectares of native marsh after low water events on other Florida lakes (Smith et al. 2004). Without substantia l changes in water regulation schedules, it is likely that Pontederia will continue to expand lakeward from its narrow bands along the shore, as happened under sim ilar conditions from 1979 to 1987 (Moyer et al. 1989). The importance of hydrological condition s to wetland control and structure is widely recognized (National Research Counc il 1996), and restoratio n of these systems must begin with hydrology (Hunt 1999). The communities established on Lake Toho following restoration may be depe ndent on consistent herbicide applications, which are increasingly regulated and dependent on currently dec lining state budgets (BIMP 2004). While restoring the variability of the nat ural hydrological regime is not possible in this or many other aquatic systems, it is possible that partial improvements to hydroperiods could be of value (Zedler 2000). For example, Lake Toho is consistently managed between 15.9 to 16.8 meters (NGVD) annually, with little to no inter-annual variability (excluding infrequent drawdowns). Y ear to year fluctuations are known to increase plant diversity, and can essentiall y double the number of vegetation types on a shoreline (Keddy and Fraser 2000). Managing lake levels with 0.5 m of inter-annual variation, for example, may reduce the amount of herbicides needed to keep robust, resilient communities from dominati ng the zone of intra-annual variation.
58 In many cases, the sequence of distur bances (including floods and droughts) and their timing, intensity, and s patial pattern have greater impac ts than single, intensive, large-scale events (Barlow and Peres 2006). Since the 1970's, intensive, managed drawdowns were conducted on Lake Toho at approximately 10 year intervals to counteract effects of water stab ilization. It is possible that incorporating year-to-year fluctuations would have a gr eater impact on littoral communi ties than infrequent, drastic drawdowns. The dense emergent communities established between these rare events proved too robust to be displaced by per iodic droughts, and appear well adapted to the degraded environmental condi tions of the lake. Additionally, large-scale, intensive disturbances may have profound impacts (Turner et al. 2003), but their outcomes ar e difficult to predict and are risky to implement. The first drawdown implemented on the lake in 1971, for example, was followed by a substantial drought, keeping lake levels below normal for nearly two years (Wegener and Williams 1974). Even if results similar to this study could be expected from large-scale, muck-removal treatments among other shallow, subtropical systems, they would likely vary considerably dependi ng on environmental conditions during and directly after application. Had this tr eatment been followed by a prolonged drought, for example, we might expect shorter hydroper iod communities to have colonized treated areas, instead of the submersed species found in this study. Ecosystem restoration should stay fo cused on key structuring processes and promoting desired ecosystem fu nctions under future conditions. The large-scale muckremoval project implemented in this study did not re-establish littoral vegetation patterns that occurred prior to water-level regul ation, but resulted in an unprecedented
59 expansion of submersed spec ies instead. While the novel submersed community that became established is preferr ed by managers over the pre -restoration habitats, it is unclear whether it will persi st under degraded environmental conditions and unchanged water regulation schedules. This study will hopefully serve as an example of the need to focus management and restoration effort s on establishing communities that will persist under current and projected condition s with minimal maintenance and effort. Even as recent studies have called for accept ing new communities instead of historical states (Seastedt et al. 2008) and promoting new approaches to managing under novel conditions (Holling 2001), we must keep t he basic premise of successful restoration efforts. Restoring or at least partially implementing key structur ing processes (Landres et al. 1999) should remain at the forefront of pattern restoration, and will minimize management contributions and costs (Mitsch and Wilson 1996). Hydrological regimes will become increasingly difficult to re store with population growth and changing precipitation patterns, and with it the temptation to maintain pattern with herbicides, species removals, and bulldozers.
60 CHAPTER 3 MUCK REMOVAL AS A TOOL FOR REST ORATION OF LITTORAL VEGETATION ON A SUBTROPICAL LAKE Introduction Much of the literature on shallow lake rest oration typically involves reducing water column turbidity and reestablishing aquatic macrophytes, either through reducing planktivore biomass, stocking piscivores (top-down approach), or reducing nutrient loads, thereby reducing phytopl ankton and increasing light penetration to the sediments (Jeppeson et al. 1990, Moss et al. 1996). These studies are based on the phenomenon of alternative stable states, where shallo w, eutrophic lakes can occur in either a degraded state (turbid water, phytoplankton-dom inated) or in a clear-water, macrophyte dominated state, under the sa me environmental conditions (Scheffer et al. 1993). Transitions between these states can occur rapidly, switching from plant to algaedominated conditions after water fluctuations (Blindow et al. 1993), plant removal (van Donk and Gulati 1995), or changes in t he food-web structure (Carpenter and Pace 1997). However, in warm temperate (Romo et al. 2005) and subtropical lakes (Bachmann et al. 2002) the correlation between high plant density and clear water is much weaker. In Florida lakes, for example, turbid wa ter can still support high macrophyte coverage, and lake restoration projects typically involve vegetation removal, rather than establishment. This approach is seemingl y at odds with much of the shallow lake restoration literature, but controlling aggressive and dense growth of littoral vegetation is a primary goal in this region. Florida lakes are subject to the same pressures facing most waterbodies near urban development, including modified lake stages (Havens 2005), elevated nutrient
61 levels (Bachmann et al. 1999), exotic specie s invasions (Smith et al. 2004), and shoreline modification (Light and Dineen 1994) Usually, such conditions result in degraded, turbid systems, but strict environm ental regulations and relatively large aquatic management budgets in the USA have im proved conditions over the last few decades; eliminating point-source pollution, im plementing fish harvest plans, promoting management of sport fish habitat, and continua lly removing invasive exotic plants (Moyer et al. 1989, Williams 2001, Allen et al 2003). Much of Florida lake management has moved beyond reestablishing macrophy tes in degraded systems, and focuses now on regulating the abundance, distribution, and co mposition of the established littoral vegetation (FFWCC 2003). There are several factors contributing to prolific littoral vegetation growth in Floridas lakes: 1) Nutrient concentrations have been reduced from their highest levels, but non-point source pollution (Havens 1995) and internal loading mechanisms (Bachmann et al. 1999) can maintain higher le vels than occurred historically, 2) Strictly regulated water schedules maintain predictable, relatively stable hydrological conditions (HDR Engineering 1989), 3) Aggressive exot ic species, including emergent (e.g. Panicum repens ), floating (e.g. Eichhornia crassipes ), and submersed (e.g. Hydrilla verticillata ) genotypes, are capable of rapid expans ion in these warm-w ater, eutrophic systems, and 4) Large area:volume ratios (lar ge photic zones) can result in up to 100% of the lake bottom capable of supporting r ooted aquatic vegetation (Canfield and Hoyer 1992). Lake restoration in this region has shi fted from vegetation establishment to removal, because without dynamic water schedules, dense emergent vegetation can
62 form floating mats around the annual low-pool elevation of the shorelines (Hoyer and Canfield 1997), possibly leading to a loss of sport fish spawning habitat, recreational access, open water area or functional lake size plant diversity, and lower oxygen levels (Moyer et al. 1989). Typical management responses include mechanical harvest of floating mats (Mallison et al. 2001), regular herbicide control of aggressive species (Schardt 1997), and implementation of man aged droughts (drawdowns), which expose organic mats to oxidation and subsidence (W egener et al. 1973). However, the shortterm benefits of these procedures (Moyer et al. 1995) and the increasing difficulty of implementing drawdowns in urban watersheds have led managers to search for new approaches, including mechanical scraping to remove muck and vegetation (hereafter referred to as muck removal). Muck removal involves scraping (with bu lldozers) vegetation and accumulated organic materials from exposed portions of the littoral zone during drawdowns, and may include regulating initial plant succession wi th herbicides. This technique has been applied to several lakes in Florida, and has been considered successful in improving short-term juvenile sport-fish abundance (Mo yer et al. 1989, Allen et al. 2003). Although no long-term faunal or vegetation m onitoring was conducted on early projects, increasingly larger applic ations have been attempted since its inception. The largest of these applications took place on Lake T ohopekaliga in centra l Florida, where approximately seven million cubic meters of shoreline vegetation and underlying soils were removed from roughly 1500 ha of the perim eter in 2004. The goal of this study was to determi ne the efficacy of this procedure by comparing preand postre storation vegetation comm unities, in both scraped and
63 unscraped treatment plots. Specifically, I tested whether muck removal resulted in different vegetation communities than a typical managed drawdown, and if so, how those responses varied by water depth and reco very rates. These results are important to lake restoration practitioners and aquatic plant managers, especially as goals shift beyond simply re-establishing vegetation and begin managing for specific habitat types. Methods and Analyses This project was conducted on Lake Tohopek aliga (hereafter re ferred to as Lake Toho), one of several large lakes in the Upper Kissimmee River Basin in central Florida, USA. Each of these lakes is successively connected and jointly regulated to collectively drain thousands of square kilometers into the Kissimmee River and ultimately Lake Okeechobee (Ch. 1, Figure 11). Toho is located near the top of this system, and covers 8,176 ha with an average depth of 2. 1 m (at maximum stage regulation of 16.75 m NGVD). This volume of water effects systems downs tream when stage levels are manipulated (e.g. Lake Okeechobee, Kissimmee Riv er), specifically since the entire chain of lakes must be lowered in order to gravitationally drawdown lakes in the northern region of the chain (HDR E ngineering 1989, Remetrix LLC 2003). Water control structures were insta lled on the Kissimmee Chain and Lake Toho in 1964 (Blake 1980), reducing the range of water level fluctuations by approximately 1.5 m to an intra-annual variability of 1.1 m (Ch. 1, Fi gure 1-2). Sewage effluent from up to three treatment facilities was discharged in to the lake from roughly 1940 to 1988, reaching a maximum of 113 m illion liters per day in 1986 (Wegener et al. 1973). The combined impacts of stage regulation and nutr ient loading were blamed for decreases in water quality and fisheries production in the la te 1960's, leading to a series of artificial dry-downs in 1971, 1979, 1987, and 2004 to improve fish habitat.
64 The largest muck removal project to date was implemented on Lake Toho in 2004. Water levels were dropped from 16. 8 m to 14.9 m from November 2003 until June 2004, during which time roughly 6.5 x 10^6 m3 of muck and vegetation were removed from roughly 1500 ha of shoreline lake bottom. Spoil was deposited at several upland sites and in 29 locations within the lake (FFWCC 2004). Water levels returned to normal by August 2004. Sample Locations Three study sites spanning 16 00 m (approx. 1 mi) of sh oreline each were located in areas with similar depth contours, vegetation communities, and shoreline use, in order to minimize interand in tra-site variation that migh t confound treatment results. Sites were selected to represent the habitat targeted by muck removal projects, which in this case was primarily a band of dense pickerelweed ( Pontederia cordata ) and floating mats of vegetation (tussocks), which occu rred roughly within the range of annual water level fluctuation. Plot sizes also had to be of functional value to avian and herpetofaunal communities for concurrent studies monitoring their responses to treatment. Each study site wa s therefore split into four treatment blocks of 400 m each, the size of which severely lim ited the amount of replicate si tes available (Figure 3-1). The spatial extent of each study plot and t he reluctance of lake managers to reserve multiple control sites throughout the lake resu lted in just three locations isolated for experimental treatment. These sites were originally structured as a randomized complete block design, with three treatments and a c ontrol randomly assigned to four blocks at each site. However, two of the planned treatments we re dropped by managers after pre-sampling was completed, so a control and treatment were randomly assign ed twice within each
65 study site. This complicated the sample design as treatments and controls were replicated within and betw een sites. Due to the efforts in minimizing inter-site variation during site selection and the dearth of replicates, I used each treatment as an independent sample to increase the sample si ze. Ideally, additional control sites would have been established on adjacent, similar la kes, but due to linked hydrologies and gravitational drawdown, other lakes of sim ilar size and region were also lowered during this application. Therefore, the design of this project was limited by cost, logisitics, and treatment restrictions, and I feel the best sa mpling protocol was established given the constraints. The maximum water depth of the plots, or the la keward extent, was delimited by the approximate water depths mechanical scrapi ng could be applied, from roughly 35 135 cm water depth at maximum lake stage.
66 N one (C ontrol) N N S /HSite 3S /H N NSite 1N S/ H NSite 2 S craped w/ Herbicide S craped w/ Herbicide N one (C ontrol) N one (C ontrol) S craped w/ Herbicide S craped w/ Herbicide 2 5 m Spray Buf fer0 1 2 3 4 5D epth classes (f eet) D epth classes (f eet) 5 4 3 2 1 02 5 m Spray Buf fer 6 D epth classes (f eet) 5 4 3 2 1 02 5 m Spray Buf fer S /H N one (C ontrol) N one (C ontrol) S craped w/ Herbicide S /H S craped w/ Herbicide N one (C ontrol)S/ H N S/H N Scraped w/ Herbicide None (Control)Depth classes (feet) 5 4 3 2 1 025 m Spray Buffer 6Scraped w/ Herbicide None (Control)S/H Figure 3-1. Location of experimental study sites on a bathymetric map of Lake Toho (0 ft, or ~ 2 m). White plots were designated controls, while blue plots were treated with mechanical mu ck removal and periodic herbicide applications. Sampling locations were stratified by three depth classes ( One = 35 cm, Two = 69 cm, Three = 103 cm) and were located on maximum slopes of 30 cm change over 30 m in distance, so as to avoid unusual shoreline contours. This was accomplished by placing 30 x 30 m grids ont o a GIS bathymetry layer (Remetrix 2003) and randomly selecting grid nu mbers from each depth category. Centroid coordinates were used to locate samples in the fiel d with a GPS (Global Positioning System) on each sampling occasion. Two locations per depth class were sampled from each treatment plot for a total of six samples per treatment, and 12 total treatment plots.
67 Aboveground vegetation wa s clipped from 0.25 m2 quadrats at each location, sorted by species, stems counted, and biomass recor ded after being oven-dried to a constant weight (70 deg C). Samples were collect ed each year during the summer (May-June) and winter (December) seasons from June 2 002 to June 2008, resulting in 13 repeated measures for each location. Soil cores were collected from the top 10 cm of substrate from each sampling location in June 2003 and June 2008, using cyli ndrical aluminum corers measuring 7 cm in diameter. Samples were placed on ice until moved to a freeze r at the University of Florida, Gainesville, FL. After being oven dried to cons tant weight, bulk densities were determined (Blake and Hartge 1986) and per cent organic content was calculated by loss on ignition (Chapman and Pratt 1961, Jacobs 1971). Lake-stage data were co llected from a South Florida Water Management District gauge located on the southern end of the lake, recording average daily water levels since January 1942. Indicators of Hy drologic Alterations (IHA) software was used to analyze pre (1942) and post waterlevel regulation (1965) periods, defined by the completion of water control structures in 1964. Both periods were analyzed for 30, 60, and 90 day maximum and minimum water levels, average monthly levels, and quartiles. Two 10 year periods before and a fter regulation, 1944 and 1992, were selected for further com parison between preand post-regulation periods. This was done to exclude some of the more extreme events in either period, including historical highs and lows experienc ed in the early 1960s, as well as managed drawdowns in the latter period. Water leve l structures were not completed on the Kissimmee Chain of lakes until 1964, but it seems reasonable that substantial changes
68 were occurring in the watershed up to severa l years prior (beginning of construction), possibly contributing to the historical highs (1960) and lows (1962) recorded just prior to completion. Rather than use these events as representative of typical pre-regulation fluctuations, the 10 year period characteriz ed by more modest events and that occurred just prior to early vegetation studies (1956) was chosen. These two periods allowed comparison of relatively stable events, ra ther than just looking at differences in maximum and minimum stage levels, which woul d also be affected by rare climatic events. Water years were defined as June through May of the following year, corresponding with the beginning of the wet season and end of the dry season, respectively. Historical vegetation data were re ferenced from management agency studies conducted in 1956, eight years prior to wate r regulation (Sincock et al. 1957), and in 1971, seven years after (Holcomb and Wegener 1971). These studies used intercept methods to calculate frequency of occurrenc e along transects perpendicular to shore, with several locations overl apping the sites in this study. These methods vary considerably from the sample techniques used here but are the only estimates of dominant species and their distributions alon g shoreline elevations prior to stage regulation. While not specif ically referenced as restorat ion targets by managers, these early data served as representatives of t he pre-stabilization, pr e-nutrient pollution, sandy substrate communities that were o ften invoked by proponent s of this project. Analyses Pre/post-restoration comparison To determine vegetation responses to the 2004 muck removal project, plant species density and biomass were summed in each quadrat fr om December 2002 and
69 June 2003 to represent pre-restorati on communities, and December 2007 and June 2008 samples as post-restorat ion representatives. Impo rtance Values (IV) were computed from these summed va lues by averaging the relative biomass and density of each species in each quadrat, and converting to a percentage value. IV = (Relative Biomass + Relative Density)/2 *100 Importance values were used to estima te species importance within a given quadrat as they are not biased towards la rge, few-stemmed species or small, numerous-stemmed species. This calculation also relativized the dataset, eliminating the need for transformations typically applied to density or biomass data that can vary by orders of magnitude between specie s and samples (McCune and Grace 2002). To reduce noise from rare species, onl y those with cumulative IVs composing 95% of the total were retained for this analysi s. Many studies el iminate species that occur in less than 1% of samples, but I used 5% of the total IV instead. This method was more representative of the actual im portance of a species throughout the sample for the same reason IVs are more repr esentative of a spec ies abundance than frequency. Analyses were limit ed to more common species as the dominant community responses to treatment were of primary inte rest, and not necessarily the loss or addition of rare species. Final matrices consist ed of 72 samples by 18 species for the preand post-enhancement periods, reduc ed from the 37 species enc ountered over the four combined sample periods. All analyses for the pre-post comparisons were performed on these matrices and unless otherwise spec ified, performed using PCORD software (McCune and Mefford 1999).
70 A hierarchical, agglomerative Cluster Anal ysis was used to find groups of similar species compositions, termed here as communi ties. Flexible beta (-0.25) linkage and Sorenson distance measures were chosen for their space conserving properties, compatibilities with each other, and their advantages with non-normal data (McCune and Grace 2002). This analysis grouped similar sample units based on species IVs, using multiple species as a basis for dec iding on the fusion of additional groups. An Indicator Species Analysis (ISA) wa s performed for two reasons: 1) to determine the optimum number of clusters for further anal ysis and 2) to define those clusters in terms of representative species. This analysis uses the proportional IV and frequency of a particular species in a particular cl uster relative to its IV and frequency in all other clusters (Dufrene and Legendre 1997) The optimum number of clusters was determined by the level of clustering that produced the most species with p-values < 0.05 in the indicator species analysis (McCune and Grace 2002). Species with low pvalues (< 0.05) and high indicator values (> 50) were used as community descriptors (cluster labels) in future analyses. A Classification and Regression Tree (CART) model (S-Plus Tree Library, Death 2002) was used to classify samples into t he groups identified by the cluster analysis using measured environmental variables alone. The Gini index was used to measure impurity (Breiman et al. 1984), cr oss-validation to select the best tree size (Vayssieres et al. 2000), and 1-Standard Error as an estima te of variation explained by the model (De'ath 2002). By building CART models for preand post-enhancement communities, species distributions could be compared ac ross environmental, spatial, treatment, and temporal gradients. Several combinations of the variables water depth, soil organic
71 content, and bulk density were used as continuous predictors, while site, treatment, and whether the sample occurred on a floating ma t were used as categorical variables. Temporal analyses In order to track changes in species biom ass over time, I combined sub-samples within each depth class and site by treatment and sample occasion, resulting in six samples over 13 time periods, each represent ing a depth class and treatment in a site. Nonmetric Multi-dimensional Scaling (NMS) ordinations were run for each site to graphically display changes in species abundances throughout the period of study (McCune and Grace 2002). Joint pl ots of correlated variables (R2 > 0.20) were overlayed onto the ordi nation diagrams, with length and di rection representative of their correlation to the axes. The sample units were color coded by treatment and consecutive samples were connected by vector s to track their movements. Within each site, ordinations were graphi cally separated by depth class for ease of interpretation. Results Soils Soil percent organic matter was plotted by treatment and time per iod, i.e. before and after the lake restoration project, usi ng all subsamples and depth stratifications (Figure 3-2). The mean perc ent organic was 33% across all samples prior to treatment, and the median was 11%. After muck remo val, the mean and median of scraped plots were both less than 2%, and were 15% and 4%, res pectively, in the control plots. After averaging across all subsamples (resulting in n = 6), Wilcoxon signed rank tests showed significant differences between the preand post-enhancement periods in both the control (P = 0.031) and the treatment plots (P = 0.016) at the 0.05 alpha level. However, I was unable to detect a signifi cant difference between treatments in the
72 amount of percent organic matter lost after restoration, usi ng a Wilcoxon rank sum test (P = 0.09). The amount of organic material lo st was more variable in the control sites, but not significantly less over all than the treated areas. Figure 3-2. A boxplot of per cent organic content in all so il cores from treatment and control plots, before (2002) and afte r (2008) the dry-down and muck removal project (before averagi ng sub-samples). Shaded boxes represent the 25th and 75th percentiles, whisker bars are the 10th and 90th percentiles, and solid circles are outliers. The means (dashed lines) and medians (solid lines) are shown within each shaded box. Vegetation Species richness was lowest in the shallo west depth class prio r to treatment, with the second and third depth class having more species in both the control and treated plots. After the dry-down and scraping, however, both plots reversed patterns and had more species in the shallowest dept h class than in the deepest (Figure 3-3).
73 0 5 10 15 20 25 234234 Pre-treatment Post-treatment Control TreatmentWater Depth (cm)# of Species 35-68 69-102 103-135 35-68 69-102 103-135 0 5 10 15 20 25 234234 Pre-treatment Post-treatment Control TreatmentWater Depth (cm)# of Species 35-68 69-102 103-135 35-68 69-102 103-135 Figure 3-3. Species richness by depth category and treatment, before and after restoration. Changes in biomass after treatment applicati on were tallied for several of the most common species (Table 3-1). These tallies we re indicative of total change, including all depth, site, and subsamples for each treatm ent type. Several species had substantial declines in biomass regardle ss of treatment, including Pontederia cordata and water lilies ( Nuphar advena, Ny mphaea odorata, and Nymphoides aquatica ). Species that had substantial increases in bi omass in control plots included Luziola fluitans Polygonum hydropiperoides Hydrilla verticillata, and Vallisneria americana The few species that benefited from t he mechanical scraping were Vallisneria, Panicum repens, Hydrilla, and Eleocharis spp. Vallisneria was not recorded prio r to the treatment, and
74 had over 1.3 kg (dry weight) totaled across the last two sample periods. P. repens had a 29% increase in biomass in scraped plot s, vs. a 35% decrease in controls. Table 3-1. Change in total dry biomass (g) fr om two preand two post-treatment sample periods of common species. Percent change could not be calculated if no biomass was recorded before treatment. Control Treat Species Biomass (g) % Change Biomass (g) % Change P. cordata 3356 66%396683% L. fluitans 1037 124%17183% Water lilies 350 72%15082% P. hydropip. 520 522%3297% P. geminatum 16 23%45500% Eleocharis spp. 63 112%2001100% H. verticillata 497 43%32068% P. repens 84 35%32029% V. americana 270 N/A1316N/A Time Series The NMS ordinations tracking species bi omass through time suggested varying stages of recovery depending on depth and site. Treatment effects were evident when the end point of the c ontrol and treatment trajectories we re 1) in different locations relative to each other, indicating differ ent compositions betw een treatments, and 2) when the treatment trajectory was stabilized in a different species space at the end of the study (new community established), and t he control trajectory was near its origin (recovery in control plots). Pearson correlations were calculated for each environmental variable and axis, and those with R2 > 0.20 were displayed on the tw o axes with the highest amount of variance explained in the three dimensiona l solution (McCune and Me fford 1999). Axes had at least a weak correlation to either soil organic matter, water depth, or both in all
75 ordinations. These correlations were portrayed with a red arrow next to the corresponding axis, with directionality indi cating increasing values of organic matter percentage or water depth. The three depth figures for eac h site were separated from one ordination to ease interpretation. The ordinations for Site 1 suggest the shallowest depth class control (dashed lines) and treatment (solid lines) plot s had recovered to near pre-treatment compositions by the end of the study, as evidenced by the proximity of the beginning and end points of the arrow trajectories (Figur e 3-4). The treatment plot in the second depth class (69 cm), however, actua lly appeared to be closer to the 2003 compositions than the control pl ot did, indicating further re covery in the treated area. The deepest samples at this site appeared to show less change overall and also seemed to be near the pre-tr eatment conditions in both control and treated plots. Site 2 ordinations suggested substantial re covery in the shallowest depth class, again with no striking difference in either trajectories or degree of recovery between treatments (Figure 3-4) The second depth class at this site, however, had a different composition in the treated pl ots by the end of the study, and an apparent recovery in the control plots. Both plots seemed to have stable communities, with very little distance (dissimilarity) between the last two sample periods. This was indicative of a strong treatment effect. Both the treatment and control plots remained different from pretreatment conditions in the deepest class, though the last sa mple in the control plot indicated recovery may be taking plac e as it moved nearer its origin.
76 Site1 Site 2 Site 3 Depth 35-68 cm (2ft) 69-102 cm (3ft)102-135 cm (4ft)Axis 2 Figure 3-4. NMS ordinations of each site with depth categories gr aphically separated. Dashed lines show control plot trajecto ries and solid lines indicate treated plots. Axes 1 and 2 are horizontal and ve rtical, respectively, for all of the graphs. Site 1: Water depth weakly negatively correlated with Axis 2 (R2 0.34) while soil percent organic matter positively correlated (R2 0.21). Site 2: Percent organic negatively correlated with Axis 1 (R2 -0.32). Site 3: Percent organic negatively correlated with Axis 1 (R2 -0.48). Site 3 had similar patterns, with the shal lowest samples having recovered in both the control and treated plots, while the second depth class showed a remarkable difference in recovery between the treatments (Figure 3-4). The scraped plot at this depth remained very dissimilar in terms of s pecies composition and biomass than it was prior to treatment, while the c ontrol plot had recovered. Agai n, this was indicative of a strong treatment effect. The deepest samples, however, showed substantially different communities in treatment and control plot s, indicating both areas remained
77 compositionally changed after restoration, regardless of whether or not they were mechanically scraped. Community Changes The four samples from the pre(December 2002 and June 2003) and posttreatment (Dec 2007 and Jun 2008) periods had 37 species, 18 of which comprised the top 95% of the cumulative importance va lues. The preand post-treatment summed datasets resulted in matrices of 72 samples by 18 species, which were divided into six communities (clusters) for the pre(3.5 % chaining, 55% information remaining) and post-restoration (2.6% chaining, 55% info rmation remaining) periods. Species with indicator values of approximately 50 or gr eater from the ISA were used to define clusters as community states (Table 32), and were labeled accordingly. The prerestoration indicators were: 1) Luziola fluitans and Panicum repens 2) Pontederia cordata and Alternanthera philoxeroides 3) Hydrocotyle spp., 4) Hydrilla verticillata 5) Lymnobium spongia and Eichhornia crassipes and 6) Utricularia spp ., Nuphar advena and Nymphaea odorata The post-treatment period also had six co mmunities identified by the cluster and ISA, which were 1) Eleocharis spp and P. repens 2) Luziola Polygonum and Paspalidium acuminatum 3) Pontederia 4) Typha spp. and Nuphar 5) Hydrilla, and 6) Vallisneria (Table 3-3).
78 Table 3-2. Indicator values of species in the pre-enhancement period (Dec 2002, Jun 2003), with values ranging 0. P_value Species Group 1 2 3 4 5 6 0.001 EICCR 49 0 0 0 0 0 0.000 LYMSP 63 0 1 2 0 0 0.000 LUZFL 0 98 0 0 0 0 0.007 PANRE 2 46 0 0 4 0 0.000 HYDSP 7 0 57 0 0 0 0.000 HYDVE 3 0 1 88 0 0 0.000 PONCO 1 11 28 0 53 0 0.001 ALTPH 3 11 13 0 46 0 0.000 UTRSP 0 0 0 0 0 94 0.001 NUPAD 0 0 0 0 0 64 0.000 NYMOD 0 0 0 0 0 48 0.697 POLHY 9 0 4 0 7 0 0.170 UNPAS 11 21 0 0 5 0 0.044 CERSP 1 0 0 20 0 32 0.047 PANHE 0 0 0 0 0 31 0.090 SAGLN 0 0 21 0 0 0 0.229 TYPSP 0 0 14 0 1 0 0.010 ELESP 0 36 0 0 0 0 Species with high indicator values are highlighted accordingly and are used as community descriptors for each group.
79 Table 3-3. Indicator values of species in the post-enhancement period (Dec 2007, Jun 2008), with values ranging 0. P_value Species Group 1 2 3 4 5 6 0.000 LUZFL 76 0 0 13 3 0 0.000 POLHY 63 0 0 3 3 0 0.000 PASAC 61 0 0 10 4 0 0.000 VALAM 0 92 3 0 0 0 0.000 HYDVE 0 5 75 0 0 9 0.000 PANRE 5 0 0 68 2 0 0.000 ELESP 4 0 1 61 9 0 0.000 PONCO 1 0 0 1 89 1 0.006 TYPSP 0 0 0 0 0 40 0.016 NUPAD 0 0 2 0 0 37 0.183 ALTPH 22 0 0 14 19 1 0.068 HYDSP 0 2 0 25 4 0 0.039 SAGLN 0 0 0 0 25 0 0.302 PANHE 4 0 0 0 11 0 0.456 CERSP 0 0 10 0 0 0 0.844 NYMOD 0 4 1 0 0 0 0.241 CHASP 0 5 16 0 0 0 0.301 UTRSP 3 0 13 0 1 0 Species with high indicator values are highlighted accordingly and are used as community descriptors for each group. Five of the six communities were predi cted by water depth, soil percent organic matter, site location, and whether the sample was categorized as a floating mat (Figure 3-5). The Eichhornia and Lymnobium community identified by the cluster and ISA analyses was not delineated at this level of pruning, and only contained 3 of the 72 samples in the analysis.
80 HYDVE (9) HYDSP (7) UTRSP_NUPAD_NYMOD (5) LUZFL_PANRE (9) PONCO_ALTPH (42) Water depth > 109 cm Site 1 Organic < 50% Water > 58 cm Sites 2 & 3 HYDVE (9) HYDSP (7) UTRSP_NUPAD_NYMOD (5) LUZFL_PANRE (9) PONCO_ALTPH (42) HYDSP HYDVE LUZFL_PANRE LYMSP_EICCR PONCO_ALTPH UTRSP_NUPAD_NYMOD HYDSP HYDVE LUZFL_PANRE LYMSP_EICCR PONCO_ALTPH UTRSP_NUPAD_NYMOD HYDSP HYDVE LUZFL_PANRE LYMSP_EICCR PONCO_ALTPH UTRSP_NUPAD_NYMODMissclassrate : Model = 0.28 Error : 0.49 Figure 3-5. CART model of pre-treatment community distribution along the measured environmental gradients. The numbers of samples in each leaf are shown in parentheses below bargraphs, which show the compositions of communities within each leaf. Species abbreviations are the first three letters of genus and first two of specific epithet (PONCO = Pontederia cordata ). The biggest difference in community types occurred at 109 cm in water depth. The most abundant community of the pre-treatment period was Pontederia and Alternanthera with 58% of all samples classified as such. This community was predicted to occur in 58 cm of water d epth at full pool, and represented the habitat type targeted by restoration efforts. The shallowest community was Luziola and P. repens occurring at < 58 cm. Deeper communiti es differed primarily by site. The Hydrilla community dominated at Site 1 at > 109 cm in water depth, while Site 2 and 3 had a Hydrocotyle community or Utricularia, Nuphar, and Nymphaea The Hydrocotyle
81 community was indicative of a floating mat, or tussock, as evidenced by the high soil organic content (> 50% ) and the fact that Hydrocotyle cannot survive in > 109 cm in water without a buoyant substrate. The best CART model for the post-treatm ent period was pruned to seven leaves, with five of the six communiti es again represented (Figure 3-6). The misclassification rate was 33%, with 53% of the variation explained (1-Standard Error). Soil percent organic matter was omitted from the final m odel, primarily to ease interpretation of treatment effects. When so il organic matter was included, several tree splits were attributed to soil differences, rather than tr eatment type. The soil differences were directly related to whether or not samples were treated, and the final groups only varied by three samples between models. I chose t he model without soils type included so as to emphasize the difference in treatment type rather than having to infer treatment from soil organic matter content. Contrary to the pre-treatment period, t he main break between community types in the post period occurred at 69 cm in water dept h, rather than 109 cm. The majority of samples in the post period were identified as Hydrilla rather than Pontederia with all control samples at > 69 cm in water depth cla ssified as such. The treated sites varied between Vallisneria or Hydrilla at > 69 cm in depth. The Pontederia group that was so prominent in the pre-treatment period was reduced to 8 samples in the post per iod, down from 58, and no longer had Alternanthera as a co-indicator. This community was classified in both treated and control areas, but was restricted to a very narrow water depth, occurring at 61 cm. There were also treatment differences in the shallowest samples, with control plots
82 having Luziola Polygonum and P. acuminatum at < 61 cm in water depth, while treated plots had an Eleocharis community at < 66 cm. Water depth > 69 cm Treatment Water < 66 cm Site 3 Control Water > 61 cm Sites 1 & 2 ELESP (7) PONCO (3) LUZFL_POLHY_PASSP (9)PONCO (5) HYDVE (22) HYDVE (9) VALAM (17)Treatment Control ELESP (7) PONCO (3) LUZFL_POLHY_PASSP (9)PONCO (5) HYDVE (22) HYDVE (9) VALAM (17) ELESP HYDVE LUZFL_POLHY_PASAC PONCO TYPSP_NUPAD VALAM ELESP HYDVE PONCO TYPSP_NUPAD VALAM ELESP HYDVE LUZFL_POLHY_PAS PONCO TYPSP_NUPAD VALAMMissclassrate : Model = 0.33 Error : 0.47Water depth < 69 cm Water depth > 69 cm Treatment Water < 66 cm Site 3 Control Water > 61 cm Sites 1 & 2 ELESP (7) PONCO (3) LUZFL_POLHY_PASSP (9)PONCO (5) HYDVE (22) HYDVE (9) VALAM (17)Treatment Control ELESP (7) PONCO (3) LUZFL_POLHY_PASSP (9)PONCO (5) HYDVE (22) HYDVE (9) VALAM (17) ELESP HYDVE LUZFL_POLHY_PASAC PONCO TYPSP_NUPAD VALAM ELESP HYDVE LUZFL_POLHY_PASAC PONCO TYPSP_NUPAD VALAM ELESP HYDVE PONCO TYPSP_NUPAD VALAM ELESP HYDVE LUZFL_POLHY_PAS PONCO TYPSP_NUPAD VALAMMissclassrate : Model = 0.33 Error : 0.47Water depth < 69 cm Figure 3-6. CART model of post-treatment community distribution along the measured environmental gradients. The numbers of samples in each leaf are shown in parentheses below each bargraph, wh ich shows the compositions of communities within each leaf. Species abbr eviations are the first three letters of genus and first two of specific epithet (PONCO = Pontederia cordata ). Discussion Muck Removal This project is the first to document li ttoral vegetation community responses to a mechanical muck removal project, particularly one of this scale. Previous studies have
83 monitored short-term (two years) changes in species frequencies in treated areas (Moyer et al. 1989), but this paper details vegetation community responses over a longer time frame (four years), establishes recovery rates, and showed how treatment effects vary by water depth. In as little as four years after restorat ion, I documented that submersed vegetation had replaced dense emer gent and floating leaf communities in treated areas, and that the newly est ablished communities showed signs of stabilization, evidenced by smaller changes in community composition between sample periods. The primary community in the z one targeted for restoration was Pontederia which dominated littoral environments in roughly 30 cm in depth prior to treatment. Typha stands were scattered at the deeper edge of the Pontederia community, and both of these species have been considered to s upport low diversity, monocultural stands that rapidly accumulate leaf litter, or or ganic sediment (Hoyer and Canfield 1997). Interestingly, this study found no consis tent change in species richness after the treatment, and an actual decrease in the to tal number of species encountered from 32 to 27. This was primarily due to a loss of ra re species or those associated with floating mats of vegetation ( Eupatorium spp Bidens laevis, Scirpus cubensis ), which have been shown to increase species richness in wetl ands by providing variably inundated substrates (Cherry and Gough 2006). Water lilies ( Nuphar, Nymphaea ) were not specifically ta rgeted by the mechanical scraping, but they were largely absent following the severa l month drawdown, presumably due to drought intolerance. These species and Typha are known to be susceptible to periodic drying events (Li et al. 2004, David 1996), and may rebound if
84 water depths remain similar to pre-treatment levels for a long enough period of time. While not evident here, some populations of water lilies were established by the end of the study outside of sampled areas. The three communities most affect ed by the restoration project, Pontederia Typha and water lilies, were completely replaced by submersed communities in all but the shallowest of treated areas. Four years after restoration, Vallisneria and Hydrilla dominated the deeper scraped sites, with Vallisneria forming dense carpets (an average of 634 plants per square meter) on the s andy substrates. This species was present prior to muck removal but was infrequent (S incock et al. 1957, Moyer et al. 1989), raising concerns about whether its rebound wil l be limited and if it will be out-competed under sustained pre-restorati on hydrologic conditions. In the shallowest of the study areas, a grassy community dominated by the exotic species P. repens replaced Pontederia which resulted in a more similar shoreline to what was described in the 1950's (Sincock et al 1957). During that period, water levels fluctuated more than two meters during some years, and these large oscillations likely favored the highly invasive exotic, P. repens This species frequently invades disturbed shorelines, and has become problematic in many of Florida's water bodies (Schardt 1997, Smith et al. 2004). The mechanical and drawdown dist urbances caused by this restoration project increased P. repens biomasses by 140% over a four year period, and expanded its distribution from < 65 cm to < 98 cm in wa ter depth. Future drawdowns or regional droughts might ext end the lakeward range of P. repens further, which is known to expand during droughts and can tolerate fl ooding depths > 1m when established (Smith et al. 2004).
85 Overall, muck removal tr eatments were most e ffective in deeper-water communities (70 cm), and had longer-las ting impacts on dense emergent or floating mat communities than submersed. For example, emer gent species like Pontederia that may have been dependent on buoyant, organic mats to survive at these depths will take longer to re-establish than su bmersed species, and no such recoveries were documented after four years in this study. Control plots were also slow to recover at these depths, but several plots had regai ned similar species compositions and biomasses within four years of drawdown. The shallowest areas (35 cm) studied, however, showed temporary treatment e ffects at best, with dense emergent communities recovering within thr ee years after muck removal. Management Implications Mechanical muck removal is used as a me thod of setting back soil or shoreline succession; by removing organic substrates and associated nutrients, re-establishing sparse communities of macrophytes, and ess entially creating a habitat more indicative of high disturbance conditions. This te chnique has been used to restore other peatlands where accumulated organics have rais ed nutrient levels and resulted in losses of diversity (Jacquemart et al. 2003) However, in a lake the size of Tohopekaliga, mechanically removing soils and associated plant communities from 1500 ha of shoreline may not realistically c hange nutrient concentrations. Half of the seven million cubic meters of material scraped from Toho was redeposited directly within the lake on spoil islands, and water column nutrient levels were unchanged two years after treatment (Hoyer et al. 2008). In reality, muck removal projects on sha llow, eutrophic lakes ar e likely to cause shifts in the compositions or distributi ons of dominant vegetation types (at least
86 temporarily), rather than lower littoral productivity or nutrient concentrations. In order to maintain a diversity of vegetation types and limit the eventual reformation of floating tussock communities, changes in lake stage r egulations must be considered. Keddy and Fraser (2000) suggested introducing inte r-annual variability in water schedules to maximize shoreline diversity, which would mo re closely mimic historical conditions on Lake Toho. Even though shoreline developm ents have lowered the maximum allowable lake stages by almost one meter, incorporat ing a half-meter of variability between water years might increase the littoral diversity. Generally, lakes with high fertility and low levels of disturbance (stable, predict able water schedules) are dominated by low diversity littoral zones, supporting large stands of nearly monocultural, highly competitive species (Grime 1979), like Pontederia Shoreline diversity can be increased by either lowering fertility or increasing dist urbance events, the latter of which could be accomplished through inter-annual stage fluct uations. Point-source pollution was eliminated on Lake Toho in the late 1980's (James et al. 1994), and furt her reductions in fertility would likely involve addressing nonpoint source runoff through watershed changes; an extremely difficult task in an urbanized and agricultural landscape. Disturbance related to flooding and drying of li ttoral communities remains a viable tool in managing shoreline vegetati on, and may reduce the need for such large-scale, intensive management projects in the future. Before applying the results of this project to other large, sha llow lakes where muck removal is being considered, several caveat s should be addressed. Lake Toho has a significant population of the submersed exotic species Hydrilla in deeper areas of the littoral zone, beyond the depths sampled in this study. Without submersed species
87 stabilizing deeper sediments, dr awdowns might increase wind-induced turbulence of the lake bottom; releasing nutrients, decreas ing water clarity, and perhaps limiting macrophyte establishment even upon reflooding (B lindow et al 1993). If a large lake already has unconsolidated sediments and deep-water portions that do not support submersed vegetation, low water levels an d the removal of shoreline vegetation may cause a shift to a phytoplankton-dominated st ate (Scheffer et al. 1993). Lake Toho had the lowest amount of subm ersed vegetation since the 1980s (BIPM 2005) immediately following this restoration, and lower water qua lity measures for nearly two years. These changes were attributed to tropical dist urbances during the reflooding phase of treatment, another caveat in t he application of these results to other systems (Ch. 4). Several conditions will likely affect the out come of mechanical muck removal projects, including lake size, water dept h, maximum fetch, nutrient concentrations, and sediment stability, any of which may affect water qua lity/turbidity and vegetation establishment in treated areas. Finally, water levels after treatment should be a major focus by management agencies, as hydrology is the primary factor in determining wetland and littoral vegetation (Pearsal 1920, Walker and Coup land 1968, van der Valk and Welling 1988). The treatments in this study were immediat ely followed by the passing of three major hurricanes, which resulted in high water events for several months following muck removal. Had lake levels risen at a slower rate or oscillated during the early recovery period, entirely different communities may have been established in treated areas (van der Valk 1981). In the case of this study, short-term res ponses were positive (though different than expected) in that the dense, robust comm unities that dominated the
88 shallow littoral zone for over 20 years were replaced with Vallisneria an important species for waterfowl (McAtee 1939, Scul thorpe 1967) and sportfish (Barnett and Schneider 1974), and which allows for bette r recreational acce ss. Longer-term monitoring is needed in order to asses the per sistence or longevity of these treatment effects.
89 CHAPTER 4 HURRICANE EFFECTS ON A LARGE-SC ALE LITTORAL HABITAT RESTORATION PROJECT Introduction The importance of water levels in r egulating wetland and littoral vegetation communities has long been recognized (Pearsa ll 1920, Segal 1971). Hydrology is typically the strongest environmental fact or controlling wetland plant community composition (Walker and Coupland 1968, van der Valk and Welling 1988), through influencing seed viability (Poiani and Johns on 1989), recruitment (Seabloom et al. 1998), and the growth and survival of adult plants (Squires and van der Valk 1992). The frequency, duration, and seasonality of floodi ng are therefore crit ical processes in determining wetland and littoral plant co mmunities (Gasith and Gafny 1990, Keddy 1983), affecting specific diversity and struct ural complexity of vegetation (Wilson and Keddy 1988, Wilcox and Meeker 1991). Lake and wetland managers have used water levels to manipulate vegetation communities for decades (e.g. Wegener 1974 ), but hydrologic schedules are often regulated by anthropogenic needs (flood control, water supply) before ecological effects are considered (Havens and Gawlik 2005). In many temperate rese rvoir lakes, water levels fluctuate too widely and reduce shoreline diversity, often resulting in unvegetated portions of littoral zone (Hill et al. 1998). In Florida, USA, where many lakes are used for flood control and water supply purposes, wa ter levels can be stabilized from their natural range and often have higher nutrient concentrations as well. A general decrease in specific diversity or structural complexity of communiti es after water-level stabilization (Keddy 1983) or eutrophicati on (Seddon 1972, Lachavanne 1985, Harper 1992) has been well documented, and such degradation is often countered with
90 periodic, infrequent drawdowns in hopes of reestablishing desirable submersed or emergent vegetation communiti es (Wegener 1974, Havens 2004) Some lakes, most recently and notably Lake Tohopekaliga (hereafte r Lake Toho) in central Florida, also have organic sediments and associated vege tation mechanically removed during the drawdown period to aid colonization of desir able species in the absence of competitive dominants (FFWCC 2003). While this effort on Toho was successful in establishing a previously undocumented, novel community in treated areas (see Chs 2 and 3), the results may have been largely influenced by natural hydrological events following the multi-million dollar project. Immediately after muck removal was comple ted and the lake began to fill, three major hurricanes passed within 65 kilometers of Toho, raising water levels 2.5 meters over a four month period to the highest lake stage since 1960 (0.5 m above maximum pool). This prolonged and deep flooding event on the heels of the re storation project likely affected early colonization and subsequent succession in treated areas. For example, without abiotic stre ssors (e.g. flooding) community composition is determined by dispersal and colonization events (Seabl oom et al. 1988), and the order of species arrival at sites has been shown to affect s pecies distributions fo r some time (Tilman 1997, Grace 1987). If the rate depth, and duration of flooding caused by hurricanes was sufficient to regulate or even eliminat e early colonizers, initial succession would have occurred after normal lake stages were restored, which may have favored the originally established communities that were removed during restoration. Without any knowledge of how these hurricanes may have affected succession and the overall outcome of this project, it would be difficult to apply these results to other projects on
91 similar lakes. The goal of this paper wa s to determine the impacts that natural disturbances may have had on this intensive restor ation effort. Specif ically, I asked: 1) Whether initial colonization compositions diffe red from those after the hurricanes, and 2) Whether the eventual dominant species were a result of the high-water events. These results are important for m anagers to understand the efficacy of this relatively new restoration approach, and serve as an important reminder that outcomes of large-scale projects may depend heavily on uncontrolla ble, external influences. Methods Mechanical muck removal began in late sp ring of 2004 after water levels reached a low of 14.9 m (NGVD), or approximately 1.0 m below normal low pool stage. The project was completed by June 2004, at which point water depths were slowly increased to regular stage levels. Standing, above gr ound, vegetation biom ass and stem density measurements were collected from 0.25 m2 quadrats in late August 2004, December 2004, and in the summer and winter of each year after until June 2008. Samples were collected late in the summer of 2004 (August) anticipating water levels would be near normal lake stages and initial succession or co lonization would have taken place in scraped areas prior to sampling. One hurricane passed over the lake in mid August (2004), which doubled the rate of water level rise over the second half of the month and brought the lake close to maximum stage by the time of the first sample (Figure 4-1). Two more major hurricanes passed in September and October of 2004, fu rther increasing lake stages to 0.5 m above full pool, the highest level recorded si nce 1960. Lake levels remained at or above maximum stage for seven months, fr om September 2004 unt il the end of March 2005. Lake stage data were collected from a South Florida Water Management District
92 gauge located at the south end of the lake (S-61H). Visual assessments of wind/wave damage were conducted after the last hurri cane passed, and other studies documented changes in water quality (Hoyer et al. 2008). 14.50 15.00 15.50 16.00 16.50 17.00 June 04 Aug 04 Oct 04Dec 04Feb 05Apr 05Lake Stage (Meters NGVD) Sample Hurricanes Sample Figure 4-1. Lake stages (meters above s ea level NGVD) following the managed dry down in summer of 2004. Dotted li nes indicate normal maximum and minimum lake regulation schedules, and the first two sample periods of early colonizers are depicted on the graph. Vegetation data were used from two ot her studies designed to assess littoral vegetation responses to the muck removal project. They were: 1) an experimental study where sampling was restricted to wa ter depth zones that were mechanically scraped, and study sites consisted of treat ed (scraped) and contro l (unscraped) plots (see Ch. 3), and 2) a lake-wide study where sampling occurred throughout the range of
93 emergent littoral zone, including areas t oo deep for bulldozers to access (145 cm) (see Ch. 2). Sampling methodologies in thes e studies differed only in that oven-dried biomasses were recorded in the former and we t biomasses in the latter. Sampling was stratified by depth category at each site, and several subsamples were collected within each stratification. Species data were summed across subsamples within each stratification, and unless otherwise noted, aver aged across sites for ea ch time period. Bare ground (empty) samples were counted individually across all stratifications. Biannual samples were also collected fo r two years prior to muck removal. Hydroperiods were calculated for sample s based on the average number of days flooded per year from 1993 thr ough 2002, a ten year period of hydrologic record immediately preceding sampling. This per iod was representative of the typical annual lake stage regulation schedules, absent artificial dry-downs that occurred in 1987 and 2004. Results Dominant species biomasses were averaged ac ross sites for the fi rst four sample periods following muck removal to evaluate initia l site colonization. In the experimental study sites, Paspalidium geminatum had the highest biomass and stem counts of all species in treated plots immediately after re flooding in August 2004 (Figure 4-2). Other prominent species in scraped areas were Pontederia cordata, Panicum repens, Polygonum hydropiperoides and Alternanthera philoxeroides These same species were recorded in the control sites but with different relative abundances, with Pontederia, Polygonum and Alternanthera dominating unscraped se ctions of shoreline (Figure 4-3). Following the hurricanes, all bi omass was severely reduced in both control
94 and treated plots, and only Pontederia and P. repens had recovered to initial levels in either plot by the fo llowing December (2005). 0 2 4 6 8 10 12 14 1234 PASGE PANRE POLHY ALTPH PONCO 10 14 12 2 0 Aug 04 Dec 04 Jun 05 Dec 05Average Dry Biomass (g) 6 4 8 Figure 4-2. Initial relative abundances of dominant species in scraped experimental plots. Dry biomasses (g) were av eraged across water depths and sites for each species. Species were coded as the first three letters of genus and the first two of specific epithet (e.g. ALTPH = Alternanthera philoxeroides ).
95 0 5 10 15 20 25 30 35 1234 PASGE PANRE POLHY ALTPH PONCO 30 35 20 10 0 Aug 04 Dec 04 Jun 05 Dec 05Average Dry Biomass (g) 15 5 25 Figure 4-3. Initial relative abundances of dominant species in control (unscraped) experimental plots. Dry biomasses (g) were averaged across water depths and sites for each species. Species were coded as the first three letters of genus and the first two of specific epithet (e.g. PANRE = Panicum repens ). The shallower, scraped secti ons of the lake-wide study sites responded similarly to the experimental treatment plots, with original colonizers including P. geminatum and Pontederia Other dominant species differed, however, with these sites having Hydrilla verticillata Eleocharis spp., and Paspalum acuminatum (Figure 4-4). Following the hurricanes, there were massive dec lines in all biomasses, and only Pontederia, P. geminatum and Hydrilla had recovered in the scraped areas by the following December (2005).
96 0 50 100 150 200 250 1234 PASAC ELESP PASGE HYDVE PONCO 100 150 200 50 0 Aug 04 Dec 04 Jun 05 Dec 05Average Wet Biomass (g) 250 Lost to High Water Figure 4-4. Initial species compositions of dominant species in scraped sections of lake-wide study sites. Wet biom asses (g) were averaged across water depths and sites for each species. Species were coded as the first three letters of genus and the first two of specific epithet (e.g. PASAC = Paspalidium acuminatum ). Total species biomass across all lake-wide sites declined considerably after the hurricanes as well, particularly in the shal lower scraped areas (Figure 4-5). However, even the deeper-water, unscraped sections of littoral zone had massive losses in biomass, reaching their lowest levels of the study between June and December of 2005.
97 012345 Average Wet Biomass (kg) per SiteJun 02 Dec 02 Jun 03 Dec 03 Aug 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08 1012 Deep Water Scraped Treat/Reflood Hurricanes Figure 4-5. Average wet biomass per site (k g) of all species in the shallow, scraped (gray) and deeper water, unscraped (black) sections of the lake-wide study sites. Standard errors of the means are shown above each bar, and the cessation of treatment and resumption of water levels is indicated in red. Most species were reduced in biomass ei ther from the mechanical removal or desiccation during the dry down, except for P. geminatum and P. acuminatum. The latter increased in biomass and abundance consi derably before the hurricanes, though it was never found at high abundance pr ior to treatment (Figure 4-6). P. geminatum was abundant throughout the study, and also had its highest biomass immediately following treatment. The vast majority of the increased biomass was from deeper-water areas where mechanical scrapi ng did not occur, where P. geminatum was already well established (Figure 4-6).
98 Deep Water Scraped Treat/Reflood 00.20.61.01.4B. Paspalidium geminatum A. Paspalidium acuminatum 010203040506070 Jun 02 Dec 02 Jun 03 Dec 03 Aug 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08 Jun 02 Dec 02 Jun 03 Dec 03 Aug 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08 Hurricanes Deep Water Scraped Treat/Reflood 00.20.61.01.4B. Paspalidium geminatum A. Paspalidium acuminatum 010203040506070 Jun 02 Dec 02 Jun 03 Dec 03 Aug 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08 Jun 02 Dec 02 Jun 03 Dec 03 Aug 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08 Hurricanes Figure 4-6. Average wet bi omass per site of A) P. acuminatum (grams) in the shallow, scraped (gray) and deeper water, unscraped (b lack) sections of the lake-wide study sites, and B) P. geminatum (kilograms). Standard errors of the means are shown above each bar, and the cessation of treatment and resumption of water levels is indicated in red.
99 Following the hurricanes and high water, P. acuminatum was never recorded in the deeper, unscraped sections of the lakewide study again, and was only occasionally found at low levels in the shallower, scraped areas. This species was found in low abundances both before and after treatment in the scraped sections of experimental plots, however, and at occasionally higher abundances throughout the study in control plots (Figure 4-7). 0246810121416Jun 02 Dec 02 Jun 03 Dec 03 A ug 04 Dec 04 Jun 05 Dec 05 Jun 06 Dec 06 Jun 07 Dec 07 Jun 08Paspalidium acuminatum 1012 Treated Plots Controls Treat/Reflood Hurricanes Figure 4-7. Average dry bi omass per site (g) of P. acuminatum in control (gray) and treated (black) sections of experimental plots. Standard errors of the means are shown above each bar, and the cessation of treatment and resumption of water levels is indicated in red. P. geminatum lost considerable biomass in t he unscraped, deeper-water sections after the hurricanes, essentially all that wa s gained during the brie f dry down. Biomass remained very similar to pre-treatment levels throughout the study per iod, however. In
100 the shallower, scraped sections, P. geminatum was at higher levels by the end of the study than occurred pre-treat ment, but still well below its deep water populations. The only species that was not recorded a fter initial colonization in the scraped plots of experiment al study sites was Polygonum hydropiperoides though its biomass increased in the control plots. All other spec ies that initially colonized the treated plots were found later in the study period. Polygonum was found throughout the study in the shallower, scraped sections of lake-wide sites. The number of empty sample s (no species) increased dramatically after the hurricanes in all sites, including the control pl ots in the experimental study (Figure 4-8). Generally, the number of empt y quadrats was low in all control and treated plots in August 2004, even though scraping had occurr ed and the lake had been reflooded. Following high water and two additional hurri canes, however, the number of empty samples increased in all sites and depth zones including the controls. Empty quadrats reached a high in all study areas roughly on e year after the hurricanes (December 2005), with the exception of t he 102 cm depth zone in experi mental plots. Both the treated and control plots had t he highest number of un-occupied samples in June of 2006, (83% and 67% respectively) two years after treatment.
101 024681012 024681012 Jun 02 Jun 03 Aug 04 Jun 05 Jun 06 Jun 07 Jun 0835-68 cm (2 ft) n = 1269-102 cm (3 ft) 102-135 cm (4 ft) 02468 0510152025303540Scraped Lake-Wide Samples n = 42 Jun 02 Jun 03 A ug 04 Jun 05 Jun 06 Jun 07 Jun 08 1012 Treated Plots Controls Treat/Reflood Hurricanes Figure 4-8. Number of empty samples in experimental plots (gray and black) and lakewide study sites. Cessation of treat ment and resumption of normal water levels are indicated in red. Re-colonization of scraped areas (after hurricanes) in the lake-wide study generally did not take place until after De cember 2006, more than two years after muck removal was completed. Total biomass of all dominant species increased dramatically beginning in June 2007, as did Vallisneria americana (Figure 4-9), the eventual dominant species in all scraped areas of the lake (Chs 2 and 3). Together with Hydrilla and P. repens these three species comprised 70% of the dominant species' total biomass from June 2007 through June 2008 (F igure 4-10) in treated areas, with P. geminatum and Pontederia rounding out the main dominants.
102 15.50 16.00 16.50 17.00 17.506/1/0412/1/046/1/0512/1/056/1/0612/1/066/1/0712/1/076/1/08 Dec 04 Dec 05Dec 06Dec 07 Jun 07 Jun 06 Jun 05Lake Stage (Meters NGVD)Total Biomass (kg) # 67 810 9111213 2 4 6 8 10 12 14 16 18 A 1 2 3 4 5 6 7 0 15.50 16.00 16.50 17.00 17.506/1/0412/1/046/1/0512/1/056/1/0612/1/066/1/0712/1/076/1/08 Dec 04 Dec 05Dec 06Dec 07 Jun 07 Jun 06 Jun 05Lake Stage (Meters NGVD)Vallisneria Biomass (kg) # 67 810 9111213 B Figure 4-9. Average daily lake stage (me ters NGVD) with horizontal dashed lines representing normal maximum and mini mum lake stages. A) Total wet biomass (kg) of all species shown in dashed black line for each sample event B) Vallisneria americana biomass (kg) shown as red dashed line.
103 0 100 200 300 400 500 600 700 800 900 1000 1234 PASGE VALAM PANRE HYDVE PONCO 800 1000 400 200 0 Dec 06 Jun 07 Dec 07 Jun 08Average Wet Biomass (g) 600 Figure 4-10. Average wet biom ass (g) of the five dominan t species at the end of the study in the lake-wide sites. Species were coded as the first three letters of genus and the first two of spec ific epithet (e.g. PASGE = Paspalidium geminatum ). Compositions in the experiment al plots were very similar to the lake-wide sites by the end of the study, dominated by Vallisneria P. repens Hydrilla and Pontederia though Eleocharis. replaced P. geminatum in the top five at these sites (Figure 4-11). The control plots were quite different, with Pontederia and Hydrilla sharing dominance with Typha Luziola fluitans and Polygonum (Figure 4-11).
104 0 20 40 60 80 100 120 1234 ELESP VALAM PANRE HYDVE PONCOA 80 100 40 20 0 Dec 06 Jun 07 Dec 07 Jun 08 60Average Dry Biomass (g) 120 0 20 40 60 80 100 120 140 160 180 200 1234 TYPSP LUZFL POLHY HYDVE PONCOB 120 160 80 40 0 Dec 06 Jun 07 Dec07 Jun 08Average Dry Biomass (g) 40 200 Figure 4-11. Average dry biom ass (g) of the five dominan t species at the end of the study in the A) treated pl ots and B) control plots of the experimental study sites. Species were coded as the first three letters of genus and the first two of specific epithet (e.g. ELESP = Eleocharis spp ).
105 The average number of Vallisneria plants per square meter during the last sample period was plotted against corresponding hydr operiods for various depth categories of samples. These hydroperiods were calcul ated based on the 10 year period of record from 1993 to 2002. Assuming lake stages retu rn to the same schedule adhered to prior to restoration, the ma jority of the depths Vallisneria was found in would be flooded less than 95% of the year (Figure 4-12). 200 400 600 800 1000 1200 Plants/m2 Hydroperiod 0 1200 200 400 1000 600 800 60-7576-9192-107108-122123-137138-152153-168169-185 0 10 20 30 40 50 60 70 80 90 100Vallisneria Plants per m2Water depth at Full Pool (cm)Average Annual Hydroperiod (1993-2002) 200 400 600 800 1000 1200 Plants/m2 Hydroperiod Plants/m2 Hydroperiod 0 1200 200 400 1000 600 800 60-7576-9192-107108-122123-137138-152153-168169-185 0 10 20 30 40 50 60 70 80 90 100Vallisneria Plants per m2Water depth at Full Pool (cm)Average Annual Hydroperiod (1993-2002) Figure 4-12. Average number of Vallisneria plants (per m2) across all samples, including experimental and lake-wide study sites. The blue line represents the average annual hydroperiod from 1993 to 2002 for each depth bin.
106 Discussion Generally, lakes with highly productiv e, stable environment s promote high biomass, broad-leaf species with slow gener ation times and the capacity for vegetative spread (Day et al. 1988, Wisheu and Keddy 1992). These communities are typically dominated by a few competitiv e species and have lower diversity littoral zones, quite similar to the Pontederia -dominated shoreline that occurred on Lake Toho prior to treatment. Lake drawdowns in Florida are generally perfo rmed to allow germination and expansion of desirable grassy species into these mono-cultural habitats, which is important to their persistence under stabilized water levels (Moyer et al. 1989). After the treatment and reflooding of Lake Toho that took place in late summer of 2004, there was substantial colonization of several gr assy and ruderal species in scraped areas, including P. acuminatum P. geminatum and Eleocharis spp. However, other species typically associated with the pre-treatm ent, degraded conditions were also present, including Hydrilla, Pontederia Alternanthera and Polygonum. The hurricanes and associated high water essentially removed the earliest colonizers; there was an average of only 3 g of dry biomass and 40 g of wet biomass collected across entire scraped study sites the fo llowing growing season (June 2005). The species described as early colonizers were typically small, newly sprouted individuals that had low biomass and hi gher densities, though there was still an additional 80% biomass reduction in all scraped areas less than a year after the hurricanes passed. All of the first col onizing species were r educed in abundance or eliminated entirely withi n that period. However, thos e same species were found again at greater abundances later in the study period, with the ex ception of one; P. acuminatum This species had the highest initia l biomass of those found in the lake-
107 wide study sites, and a density of roughly 60 plants per square meter. After the hurricanes that density was reduced to 9 per square meter, and it was rarely collected from those sites after. This species was also found in the ex perimental plots, but infrequently and with low abundance relative to common species. There was no increase in densities or biomass in treated plots like there was in t he lake-wide studies, but its biomass was still reduced after the hurricanes. However, it was found later in both control and treated plots at the same low same low frequency a nd abundance that it had pr ior to treatment. This indicates that the increases recorded in the lake-wide sites were somewhat of a patchy event, and indeed occurred in only three of the five sites. At least one other study documented an increase in P. acuminatum following low water events in Florida lakes (Havens 2005), but this is generally not a dominant shoreline species in the region. Several local wetland flora books and regional plant I.D. websites do not list the species, further suggesting its local infrequency. While it is unknown whether this specie s could have remained at least locally abundant in shallow, scraped secti ons of shoreline if the hurricanes had not passed, it seems probable that other regi onally dominant species, spec ifically invasive exotic grasses like P. repens would have replaced it. By the end of the study P. repens dominated all scraped plots in bot h the experimental and lakewide study areas. This species typically colonizes disturbed sites and is very problematic in many of Florida's water bodies, having displaced thousands of acres of native marsh in other systems following droughts or low water events (Smith et al. 2004). Given the potential for P. repens to invade and persist in disturbed areas, as well as the fact it has been common
108 on Lake Toho since the earliest vegetation st udies (Sincock et al. 1957, Holcomb and Wegener 1971, Moyer et al. 1989) it seems unlikely that P. acuminatum a relatively infrequent species in the region w ould have remained abundant for long. One of the biggest impacts of the hurricanes may have taken place at depths beyond where scraping took place, where P. geminatum was the most common and abundant species. Like most grasses, this species germinates during low water periods (Johnson et al. 2007.) and is typically found on Toho and other regional lakes at depths concurrent with previo us drawdowns. It was predicted that P. geminatum would expand shoreward after organic substrates and dense competitors were removed (see http://aquat1.ifas.ufl .edu/guide/laketoho.html for project details), and P. geminatum was one of the few species to substantially in crease in biomass i mmediately after the treatment (primarily in deeper, unscraped areas). It was also one of the early colonizers of treated shorelines and one of the first to recover following hurricane damage. However, the biomass increase from germinat ion in deeper areas was immediately lost after the hurricanes, presumably because t he dense, new growth could not withstand the high winds or water levels. Visual obs ervations following the hurricanes confirmed the biomass reductions shown in the samples, as most shorelines had P. geminatum wrack deposited at the high water line. Even after a nearly 50% reduction in biomass, P. geminatum abundance was still well within pre-treatment levels direct ly after the hurricanes and throughout the remainder of the study. There likely wo uld have been much higher densities and possibly an expansion into previously uninh abited areas if the newly germinated plants could have survived rising water levels. For example, most of t he lily pad communities
109 ( Nuphar advena and Nymphaea odorata ) were largely eliminated from the scraped and deeper water sections of shoreline, presum ably due to desiccation and rapid water level increases (Ch. 2). It seems reas onable that the dry down expansion of P. geminatum into those areas previously occupied by dense floating leaf communities could have remained for some time. Overall, the initial composition of early site colonizers did not appear to change much after the hurricanes. Ea rly growth was severely limited due to high water, but the initially dominant species all recovered biom ass after successive years of normal lake stages. The compositions found after nearly four years of post-tr eatment succession did not begin to take form until after the pr e-hurricane colonizers had recovered. For example, P. geminatum, Pontederia, El eocharis, Hydrilla, Polygonum and Alternanthera were all dominant species initially, and were again dominant two growing seasons after the hurricanes passed (June 2006). Beginning in December of 2006, Vallisneria and P. repens became increasingly dominant. In other words, the final composition documented at the end of the study formed afte r the initial colonizers were already in place. It seems reasonable that if water leve ls had followed normal lake regulation schedules after the treatment, t he same pattern of site colo nization would have taken place, but most likely a year or two sooner There most likely would have been more P. geminatum particularly in the deeper unscraped areas, but given that fi erce competitors like Pontederia and P. repens were among the early colonizers of scraped areas, it seems that the same succession pattern would have taken plac e at a more rapid pace. These were the results in a previous, small-scale muck removal on Lake Toho, where
110 within two years of treatment Pontederia, Alternanthera Hydrilla P. repens and Eleocharis were again among the most dominant species (Moyer et al. 1989). These results also fit wetland communi ty succession models, where, following drawdowns, perennial communities rapidly rees tablish once original water levels are returned (Seabloom et al. 2001). Conversely, th is rate of establishment has been linked to the level of flooding (deviation from normal) during the recovery period. If the normal annual lake stages had been restored immediately following rest oration, it seems likely that the previously dominant s pecies would have had the advantage. The succession studies and models predicti ng rapid recovery assume adequate propagule availability, however, and this may not be the case in this study. When original communities fail to reestablish, disp ersal capabilities or propagule availability are typically a limiting fa ctor (Ellison and Bedford 1995) or conditions may have changed such that the original cohorts can no longer persist at the site (Turner et al. 1998). The restoration proj ect on Lake Toho probably falls somewhere in between these two scenarios, as sediment characteri stics were altered, water depths were slightly increased, and the seedbank was at l east severely disturbed, if not relocated entirely. However, half of the sediment and/or vegetation that were removed form the shoreline was deposited back into the sha llow littoral zone on spoil islands, and only about half of the emergent littoral zone wa s scraped (the shallower half, 45 cm). While seed availability and vegetative regen eration were surely limited following the treatment, it seems reasonable t hat all of the original spec ies were still within dispersal range of the scraped sites. Given that adult plants were removed, initial succession would most likely have been driven by germinat ion and dispersal limitations instead of
111 clonal or rhizomatous spreading. However, the close proxim ity of available propagules, similar species pool, and the inevitable return of stable hydrologic conditions will most likely favor reestablishment of t he original communities eventually. The major surprise in this project wa s the invasion and at least short-term persistence of a previously rare species on Lake Toho, Vallisneria This submersed species rapidly expanded in areas of anot her regional lake (O keechobee) following low water periods, and apparently is capable of displacing some submersed species under the right conditions (Havens et al. 2004). Vallisneria and most other species did not rapidly expand until unusually low winter water levels in 2006, after which the current dominant species increased dramatically. Th is may have been due to high turbidity and low water quality that persisted in the lake for up to two years following the hurricanes (Hoyer et al. 2008), presumably from fl ushing surrounding agricultural lands and a massive decline in macrophyte abundanc e after the treatment. While Vallisneria is fairly tolerant of low light conditions, it has higher light requirements for early growth than it does for germination (Kimber et al. 1995). It is possible that extended low winter water levels during 2006 aided in the expansion of Vallisneria from deeper water, unscraped habitats. The communities found after f our years of succession were likely a result of recent hydrological conditions, as well as dispersa l capabilities. Two of the top four species were invasive exotics, apparently displacing slower growing, broa d leaf emergents like Pontederia for the time being. However, the dominance of submersed species, both Vallisneria and Hydrilla in scraped areas, is likely due to higher water levels in general following the treatment. More t han half of the areas occupied by Vallisneria have an
112 annual hydroperiod < 95% under current lake regul ations, elevations typically occupied by emergent species. Had water levels be en held below maximum or at minimum lake stage for a full growing season after the restor ation, there likely would have been much less submersed species and a more rapid expansion of aggressi ve emergents, like Pontederia and P. repens While prolonged high water and in creased turbidity from the hurricanes undoubtedly affected early colonize rs, the communities found at the end of this study seemed to be a result of several growing season hydrologies. Keeping water levels lower for a full growing season afte r the enhancement projec t likely would have promoted more P. geminatum in scraped areas, but substantially less Vallisneria. As many studies have shown, the hydrologic regime during colonization/germination of wetland soils will has a large impact on init ial community composition (van der Valk 1981, Seabloom et al. 1998), but ultimately species distributions will be based on flooding tolerances (Squires and van der Va lk 1992). Without changes to annual lake schedules, the communities that succeed afte r such large-scale lake restoration projects will likely be very similar to pre-treat ment compositions once initial dispersal limitations are overcome.
113 CHAPTER 5 DISCUSSION Review This study was the most thorough documentat ion of littoral vegetation responses to mechanical muck removal, particularly on this scale of application. Previous studies monitored short-term (two y ears) changes in species frequencies in scraped areas after treatment (Moyer et al. 1989), but this paper details vegetation changes over a longer time frame (four years), includes pre-treatme nt descriptions, and monitors shifts in distributions along a water depth gradient. The overall success of this restoration project was mixed, in that specified compositions were not achieved, but desir ed habitat characteristics were. The FFWCC predicted that over time healthy stands of knotgrasses (more commonly known as P. geminatum ), bulrushes ( Scirpus californicus ), and eelgrass ( Vallisneria ), among other desirable species, would rebound from the seedbanks as wo rkers managed against problematic species (http://aquat1.ifas.ufl .edu/guide/laketoho.html ). The primary community, or p roblematic species target ed by the restoration was Pontederia which I found dominated in roughly 30 cm in water depth prior to treatment, with Alternanthera identified as another strong indi cator for this community. Shoreward of this gr oup was a mixture of Luziola, P. repens, and Eleocharis all of which have been found at varying frequencies since the 1950s (Sincock et al. 1957, Holcomb and Wegener 1971, Moyer et al. 1989). The deepest water edge of the Pontederia/Alternanthera community was bordered by occasional floating mats, including Hydrocotyle Lymnobium, and Eichhornia, indicating an abrupt shift from dense emer gent communities to deeper-water habitats.
114 These were the communities believed to fo rm an organic, floatin g barrier to fish attempting to access shallow spawning areas (Moyer et al 1989). Based on a 10 yr period of stage records (1993) the lakeward extent of the dominant Pontederia zone extended just beyond the annual minimum water depths. Typha stands were scattered along the deepest edge of this community, as cattails generally benefit from several cent imeters of flooding for germination, but not complete exposure during annual lows (J ohnson et al. 2007). Beyond the 100% hydroperiod zone there were wate r lily communities, primarily Nuphar and Nymphaea while Hydrilla and P. geminatum dominated the deepest portions of the emergent littoral zone. Soil organic content was highly variable bef ore lake restoration, ranging anywhere from 2% to 96% in the top 10 cm of substrate, with a mean of 33% across plots and a median of only 11%. These numbers dropped to barely detectable levels four years after treatment (< 2%), but were even significantly reduced in control plots (mean 15%, median 4%). While peat depths were not ac tually measured, the soil corer generally reached a sand or silt subs trate underneath the organic la yer while sampling, which often occurred before 10 cm and made core retrieval more difficult. There were few instances where the amount of peat or organic materi al exceeded 10 cm, though samples from a cove and floating mat exceeded 20 cm on a few occasions. Overall, the objective of removing exis ting plant monocultures and associated organic material was achieved, though there wa s only a slight rise in the effective number of species in scraped areas after restoration (5.3 to 6.2). However, given the loss of water lily communities from deeper, un scraped areas, the total effective species
115 for the emergent littoral zone dropped from 7.9 to 5.5 after restoration. It is unclear what role the hurricanes may have played in eliminating deeper water Typha and water lily communities. The dense Pontederia community was prim arily replaced by P. repens in the shallowest sections of shoreline (< 98 cm), and by Vallisneria in the deeper areas. Historically, P. repens was the most frequently occurri ng species along study transects, occupying nearly the same elevations Pontederia did due to greater shoreline fluctuations (Sincock et al. 1957). However, Vallisneria was very infrequent in early studies, and Lake Toho was described as havi ng too low an alkalinity to support much submersed vegetation (Sincock et al. 1957). While the replacement of Pontederia with P. repens in shallow water does more closely resemble pre-regulation (histo ric?) shorelines, the expansion of Vallisneria was unprecedented in this system. Havens et al. (2004) documented a recovery of Vallisneria following low water on Lake Okeechobee, bu t this presumably occurred from existing seedbanks and was a ccompanied by a rebound of Potomageton and Hydrilla as well. The seedbank was largely remov ed or at least severely disturbed on Toho following muck removal, and Vallisneria was rarely seen prior to treatment (Moyer et al. 1989), indicating a possible rapid expansion fr om isolated, deeper-water patches. This suggests that while Vallisneria remained on the lake for over 50 years, it was apparently a poor competitor under both historic and pr e-treatment conditions, never being more than infrequently recorded. The question is, wa s establishment the only factor limiting its distribution, and can it continue to occ upy large areas under the same environmental conditions and species associations as before?
116 P. geminatum was predicted to be the major benef actor of this restoration, including expansion into areas previously occupied by Pontederia (http://aquat1.ifas.ufl .edu/guide/laketoho.html ). This species began establishing in scraped areas immediately after muck remo val but was reduced in abundance following major hurricanes in 2004. There were substantial increase s in deeper-water P. geminatum communities from the drawdown as we ll, but those gains were also lost during the hurricanes. While water levels during initial colonization undoubted ly affected establishment rates and compositions, P. geminatum was still one of the fe w species present in scraped areas even a year after the hurrica nes passed. While never dominant, P. geminatum remained one of the top five species in terms of total biomass in treated areas at the end of the study. This suggests that while P. geminatum was inhibited by the rapid water rise and prolonged flooding from the hurricanes, it faired better than other colonizing species and still was unab le to dominate the scraped areas as predicted. Overall, the short-term responses of this restoration were t hat the dense, robust communities that dominated the shallow litto ral zone for over 20 years were replaced primarily with Vallisneria an important species for waterfowl (McAtee 1939, Sculthorpe 1967) and sportfish (Barnett and Schneider 1974) and which permits better recreational access to restored areas. Whether Vallisneria can remain dominate at these depths and/or at what cost is unknow n, as half of the zone occupied by this submersed aquatic was exposed annually from 1993 to 2002, and the invasive exotic Hydrilla is known to displace this species in fertile environments (Van et al. 1999). Hydrilla biomass
117 increased 58% after the restor ation, and though this was prim arily in deeper areas than where Vallisneria occurred, it still suggests the pot ential for invading scraped areas without continual intervention. As such, the Vallisneria community is likely to be compressed from emergent competition in areas with < 100% hydroperiod, and from Hydrilla and possibly recovering water lily communi ties at the deeper end of its gradient. Management Implications The muck removal treatment applied to Lake Toho was the largest and most intensive removal event to date, and was used as a substitute for the loss of historic flood/drought cycles that would have natur ally flushed/oxidized accumulated organic materials or floating mats during such event s. This unique approach was an attempt to reset decades of succession by removing the species, seedbanks, and soils associated with degraded shorelines. Recolonization was regulated with selective herbicide applications in an effort to force historical patterns on the sandy, exposed shorelines without the historical processes that maintai ned the system originally. After four years of post-treatment monito ring, these efforts appeared to have mixed results. While the objective of establishing spar se emergent habitat was not realized, the communities that developed along the treat ed shorelines did have similar desirable attributes as the target habitat. Organic so ils were replaced with sandy substrates, the submersed communities that were established provide important habi tat for fish (Barnett and Schneider 1974) and waterfowl (McAtee 1939, Sculthorpe 1967), and the removal of floating mats and dense emergent vegetati on allowed for better recreational access. When system degradation cannot be reversed at an acceptable cost, managers should strive for the biotic structure and ecosyst em services desired by stakeholders, while promoting communities that ar e both feasible and resilient (Seastedt et al. 2008). This
118 project did accomplish the form er, but the resilience or lo nger-term establishment of a submersed community at these lake elevat ions is doubtful. A regional drought could expose most of the shoreline occupied by the newly established submersed community, which would allow rapid invasi on by emergent species like Pontederia and especially P. repens (Smith et al. 2004). Without substantia l changes in water regulation schedules, it is likely that Pontederia will continue to expand lakeward from its narrow bands along the shore, as happened from 19 79 to 1987 (Moyer et al. 1989). The importance of hydrological condition s to wetland control and structure is widely recognized (National Research Counc il 1996), and restoratio n of these systems must begin with hydrology (Hunt 1999). The communities established on Lake Toho following restoration may be dependent on cons tant herbicide applications, which are increasingly regulated and dependent on currently dec lining state budgets (BIMP 2004). While restoring the variability of the nat ural hydrological regime is not possible in this or many other aquatic systems, it is possible that partial improvements to hydroperiods could be of value (Zedler 2000). For example, Lake Toho is consistently held between 15.9 and 16.8 meters NGVD annua lly, with little to no inter-annual variability. Year to year fluctuations are known to in crease plant diversity, and can essentially double the number of vegetation types on a s horeline (Keddy and Fraser 2000). Managing lake levels with 0.5 m of inter-annual variation, for example, may reduce the amount of herbici des needed to keep robust, resilient communities from dominating the zone of intra-annual variation. Ecosystem restoration will likely become increasingly complex as more systems are exposed to novel conditions, but s hould continue to focus on key structuring
119 processes and promoting desired ecosystem functions under future conditions. Even as recent studies have called for accepting ne w communities instead of historical states (Seastedt et al. 2008) and promoting new approaches to managing under novel conditions (Holling 2001), the basic premise of successful restoration attempts should not be abandoned. Restoring or at least par tially implementing key structuring processes (Landres et al. 1999) should remain at the forefront of pattern restoration, and will minimize external efforts and cost s (Mitsch and Wilson 1996). Hydrologic regimes will become increasi ngly difficult to restore with population growth and changing climates, and with it the temptation to maintain pattern with herbicides, species removals, and bulldozers, in stead of water level manipulation. Before applying the results of this project to other large, sha llow lakes where muck removal is being considered, several caveat s should be addressed. Lake Toho has a significant distribution of Hydrilla populations in deeper areas of the littoral zone, beyond the depths sampled in this study. Wit hout submersed species stabilizing deeper sediments, drawdowns might increase windinduced turbulence of the lake bottom; releasing nutrients, decreasing water clarity, and perhaps limiting macrophyte establishment even upon refloodi ng (Blindow et al 1993). If a large lake already has unconsolidated sediments and deep-water por tions unsupportive of submersed vegetation, low water levels and the removal of shoreline vegetation may cause a shift to a phytoplankton-dominated st ate (Scheffer et al. 1993). Lake Toho had the lowest amount of s ubmersed vegetation since the 1980s (BIPM 2005) immediately following this rest oration, and lower water quality measures for nearly two years. These changes were attributed to tropical disturbances during the
120 reflooding phase of treatment, which is another important consideration when applying large-scale habitat modifications. Several c onditions will likely affect the outcome of mechanical muck removal projects, including lake size, water depth, maximum fetch, nutrient concentrations, and sediment stab ility, any of which may affect water quality/turbidity and subsequent vegetation establishment in treated areas. Finally, water levels after treatment shoul d be a major focus in these restoration efforts, as they are instrumental in in fluencing seed viability (Poiani and Johnson 1989), recruitment (Seabloom et al. 1998), and the gr owth and survival of adult plants (Squires and van der Valk 1992) on the scraped shorelines The hurricanes that occurred during this study resulted in a rapid increase of wa ter levels that were held at maximum pool for nearly seven months. Had water returned at a slower rate or oscillated during the early recovery period, diffe rent communities may have been established (van der Valk 1981, Weiher et al. 1996). Longer-term monitoring is recommended after su ch projects, as the treated areas were only beginning to stabilize by the end of this study. Several of the dominant species were still consistently increasing in biomass over the last few samples, including the target species Pontederia and its large-scale replacements, P. repens and Vallisneria There is still valuable informati on about the long-term persistence of establishing communities or recovery of undesirable communi ties that has not yet been collected. Without adequate monitoring of m anagement efforts, it will be impossible to learn from their l ong-term effects.
121 APPENDIX A SPECIES LIST Table A. Common species sampled over the period of study. Species Code Scientific Name Common Name ALTPH Alternanthera philoxeroides Alligator weed BACCA Bacopa caroliniana Lemon Bacopa BIDLA Bidens laevis Burrmarigold BRAMU Bracharia mutica Para grass CENAS Centella asiatica Coinwort CERAT Ceratophyllum spp. Coontail CHASP Chara spp. Musk grasses CYPSP Cyperus spp. Sedges DIOVI Diodia virginiana Buttonweed EICCR Eichornia crassipes Water hyacinth ELESP Eleocharis spp. Spikerushes HABRE Habenera repens Water-spider orchid HYDSP Hydrocotyle spp. Pennywort HYDVE Hydrilla verticillata Hydrilla LEESP Leersia spp. Cut grasses LUDAR Ludwigia arcuata Piedmont primrose LUDSP Ludwigia spp. Ludwigia/Water Primrose LUZFL Luziola fluitans Water grass LYMSP Lymnobium spongia Frog's bit NAJGU Najas guadalupensis Southern naiad NELLU Nelumbo lutea Water lotus NITSP Nitella spp. Stoneworts NUPLU Nuphar advena Spatterdock NYMAQ Nymphoides aquatica Banana lily NYMOD Nymphaea odorata Fragrant water lily PANHE Panicum hemitomon Maidencane PANRE Panicum repens Torpedo grass PASGE Paspalidium geminatum Egyptian paspalidium (commonly knot grass or Kissimmee grass) PASAC Paspalum acuminatum Canoe grass POLDE Polygonum densiflorum Smartweed POLHY Polygonum hydropiperoides Wild water-pepper PONCO Pontederia cordata Pickerel weed RHYNSP Rhyncospora spp Beakrushes SAGLN Sagittaria lancifolia Duck potato SAGLT Sagittaria lattifolia Arrowhead
122 Table A. continued. Species Code Scientific Name Common Name SCICA Scirpus californicus Giant bulrush SCICU Scirpus cubensis Bulrush SESPU Sesbania punicea Purple Sesban TYPSP Typha spp. Cattails UTRSP Utricularia spp. Bladderworts VALAM Vallisneria americana Eelgrass Nomenclature follows that of Tobe et al. 1998.
123 LIST OF REFERENCES Abrahams, C. 2008. Climate c hange and lakeshore conservation: a model and review of management techniques. Hydrobiologia 613:33 43. Allen, M. S., K. I. T ugend, and M. J. Mann. 2003. Largemouth bass abundance and angler catch rates following a habitat enhan cement project at Lake Kissimmee, Florida. North American Jour nal of Fisheries Management 23:845 855. Anderson, R. C., J. E. Scw egman, and M. R. Anderson. 2000. Micro-scale restoration: a 25-year history of a southern Illinoi s barrens. Restoration Ecology 8:296 306. Bachmann, R. W., M. V. Hoye r, and D. E. Canfield. 19 99. The restoration of Lake Apopka in relation to alternative stable states. Hydrobiologia 394:219 232. Bakker, J. P., and F. Berends e. 1999. Constraints in the restoration of ecological diversity in grassland and heathland co mmunities. Trends in Ecology and Evolution 14:63 68. Barlow, J., and C. A. Peres. 2006. Cons equences of cryptic and recurring fire disturbances for ecosystem structure and bi odiversity in Amazonian forests. Pages 225 240 in W. F. Laurance and C. A. Peres, edi tors. Emerging threats to tropical forests. University of Chicago Press, Chicago. Barnett, B. S., and R. W. Schneider. 1974. Fish populations in dense submersed plant communities. Hyacinth Control Journal 12:12 14. Blackman, W. F. 1973. History of Orange Count y. Mickler House P ublishers, Chulota, FL. Blake, G. R., and K. H. Hartge. 1986. Bulk density. Pages 464 475 in A. Klute, editor. Methods of soil analysis: Part 1. Physica l and mineralogical methods. American Society of Agronomy, Madison, WI. Blake, N. M. 980. Land into water water in to land. University Presses of Florida, Tallahassee, FL. Blindow, I., G. Anderson, A. Hargeby, and S. Johansson. 1993. Long-term pattern of alternative stable states in two shallo w eutrophic lakes. Freshwater Biology 30:159 167 Breiman, L., J. H. Friedman, R. A. Ols hen, and C. J. Stone. 1984. Classification and regression trees. The Wads worth Statistics/Probability Series, Chapman & Hall, New York, NY. Bureau of Invasive Plant Management. 2004. St atus of the Aquatic Plant Maintenance Program in Florida Public Waters. Annual Report to Florida State Legislature. Department of Environmental Protection.
124 Bureau of Invasive Plant Management. 2005. St atus of the Aquatic Plant Maintenance Program in Florida Public Waters. Annual Report to Florida State Legislature. Department of Environmental Protection. Canfield, D. E., and M. V. Ho yer. 1992. Aquatic macrophyte s and their relation to the limnology of Florida Lakes. Final Repor t. Bureau of Aquatic Plant Management, Tallahassee, FL. Carpenter, S. R., and M. L. Pace. 1997. Temporal variat ion in a paleolimnological record arising from a trophic cascade. Ecology 72:277 285. Chapin III, F. S., A. L. Lov ecraft, E. S. Zavaleta, J. Ne lson, M. D. Robards, G. P. Kofinas, S. F. Trainor, G. D. Peterson, H. P. Huntington, and R. L. Naylor. 2006. Policy strategies to address sustainability of Alaskan boreal forests in response to a directionally changing climate. Pr oceedings of the Na tional Academy of Sciences USA 103:16637 16643. Chapman, H. D., and P. F. Pr att. 1961. Methods of analysis for soils, plants and waters. University of California, Riverside, CA. Cherry, J. A., and L. Gough. 2006. Tempor ary floating island formation maintains wetland plant species richness: The role of the seed bank. Aquatic Botany 85:29 36. Cissel, J. H., F. J. Swanson, and P. J. Weisberg. 1999. Land scape management using historical fire regimes: Blue River, Oregon. Ecological Applications 9:1217 1231. Davis, S. M., and J. C. Ogden. 1997. Everglades: the ecosyst em and its restoration. St. Lucie Press, Boca Raton, FL. David, P. G. 1996. Changes in plant communities relative to hydrologic conditions in the Florida Everglades. Wetlands 16:15 23. Day, R. T., P. A. Keddy, J. McNeil, and T. Carleton. 1988. Fertility and disturbance gradients: a summary model for ri verine marsh vegetation. Ecology 69:1044 1054. De Angelis, D. L., L. J. Gro ss, M. A. Huston, W. F. Wolf f, D. M. Fleming, E. J. Comiskey, and S. M. Sylvester. 1998. Landscape modeling for Everglades ecosystem restorati on. Ecosystems 1:64 75. Death, G. 2002. Multivariate regression tr ees: a new technique for modeling speciesenvironment relationships. Ecology 83:1105 1117. Diehl, S. 1992. Fish predation and benthic comm unity structure: The role of omnivory and habitat complexity. Ecology 73:1646 1661.
125 Dobson, A. P., A. D. Bradshaw, and A. J. Ba ker. 1997. Hopes for the future: restoration ecology and conservation biology. Science 227:515 522. Doren, R. F., and L. D. Whiteak er. 1990. Effects of fire on di fferent size individuals of Schinus terebinthifolius Natural Areas Journal 10:107 113. Dufrene, M., and P. Legendre. 1997. Species a ssemblages and indicator species: the need for a flexible asymmetrical approach. Ecological Monographs 67:345 366. Ellison, A. M., and B. L. Bedford. 1995. Response of a wetland vascular plant community to disturbance: a simulati on study. Ecological Application 5:109 123. Fan, A., and S. Lin. 1984. Water budget for upper Kissimmee chain of lakes. Technical Memorandum. South Florida Water management District, West Palm Beach, FL. Florida Fish and Wildlife Conservati on Commission. 2003. 2004 Lake Tohopekaliga habitat enhancement project: a fis hery management program. Fresh Water Fisheries Division, FFWCC. Florida Fish and Wildlife Conservation Commission. 2004. Kissimmee Chain of Lakes Highlights, August 13, 2004. Aquatic Habitat Conservation and Restoration Section, Kissimmee, Florida, USA. Gasith, A., and S. Gafny. 1990. Effects of water level fluctuation on the structure and function of the littoral zone. Pages 156 in M. M. Tilzer and C. Serruya, editors. Large Lakes: Ecological structure a nd function. Science-Tech Publication. Madison, Wisconsin Giudicelli, J., and M. Bournaud. 1996. Invertebr ate biodiversity in land-inland water ecotonal habitats. Pages 143 160 in J.-B. Lachavanne and R. Juge, editors. Biodiversity in land-inland water ecotones. Parthenon Publishing Group, New York, NY. Grace, J. B. 1987. The im pact of preemption on the z onation of two Typha species along lakeshores. Ecological Monographs 57:283 303. Grime, J. P. 1979. Plant strategies and v egetation processes. Wiley, Chichester. Harper, D. 1992. Eutrophication of freshwaters: principles, problems and restoration. Chapman & Hall, London, UK. Havens, K. E. 1995. Secondary nitrogen limitati on in a subtropical lake impacted by non-point source agricultural polluti on. Environmental Pollution 89:241 246. Havens, K. E. 2005. Aquatic vegetation and largemouth bass population responses to water-level variations in Lake Ok eechobee, Florida (USA). Hydrobiologia 539:225 237.
126 Havens, K. E., N. G. Aumen, R. T. Jame s, and V. H. Smith. 1996. Rapid ecological changes in a large subtropical lake under going cultural eutrophication. Ambio 25:150 155. Havens, K. E., B. Sharfstein, M. A. Brady, T. L. East, M. C. Harwe ll, R. P. Maki, and A. J. Rodusky. 2004. Recovery of submerged pl ants from high water stress in a large subtropical lake in Florida, USA. Aquatic Botany 78:67 82. Havens, K. E., and D. E. Gawlik. 2005. Lak e Okeechobee conceptual ecological model. Wetlands 25:908 925. HDR Engineering. 1989. Technical report of the development of a Surface Water Improvement and Management (SWIM) pl an for Lake Tohopekaliga/East Lake Tohopekaliga. Final Report. South Flori da Water Management District, West Palm Beach, FL. Hill, M. 0. 1973. Diversity and evenness: a unifying notation and its consequences. Ecology 54:427 432. Hill, N. M., P. A. Keddy, and I. C. Wisheu. 1998. A hydrologi cal model for predicting the effects of dams on the shoreline vegetati on of lakes and reservoirs. Environmental Management 22:723 736. Hobbs, R. J., and D. A. Norton. 1996. Toward s a conceptual framework for restoration ecology. Restoration Ecology 4:93 110. Hobbs, R. J., S. Arico, J. Aronson, J. S. Baron, P. Bridgewater, V. A. Cramer, P. R. Epstein, J.J. Ewel, C. A. Klink, A. E. Lugo, D. Norton, D. Ojima, D.M. Richardson, E. W. Sanderson, F. Valladares, M. Vila R. Zamora, and Martin Zobel. 2006. Novel ecosystems: theoretic al and management aspects of the new ecological world order. Global Ecol ogy and Biogeography 15:1 7. Holcomb, D., and W. Wegener. 1971. Hydrophytic ch anges related to lake fluctuation as measured by point transect s. Proceedings of the 25th Annual Conference of the Southeastern Association of Game and Fish Commissioners. Colombia, SC Holling, C. S. 2001. Underst anding the complexity of econom ic, ecological, and social systems. Ecosystems 4:390 405. Holm, L.G., L.W. Weldon, and R. D. Blackburn. 1969. Aquatic weeds. Science 166:699 709. Hoyer, M. V., and D. E. Canfield. 2007. Aquatic plant management in lakes and reservoirs. North American Lake Managemen t Society, Madison, WI and Aquatic Plant Management Society, Lehigh, FL
127 Hoyer, M. V., R. W. Bachm ann, and D. E. Canfield, Jr. 2008. Lake management (muck removal) and hurricane impacts to the trophi c state of Lake Tohopekaliga, Florida. Lake and Reservoir Management 24:57 68. Hunt, R. J., J. F. Walker, and D. P. Krabbenhoft. 1999. Characterizing hydrology and the importance of ground-water discharge in natural and constructed wetlands. Weltands 19:458 472. Hunter, M. L. 1993. Natural re regimes and spatial models for managing boreal forests. Biological Conservation 65:115 120. Jacobs, H. S. 1971. Soils laboratory exercise source book. American Society of Agronomy, Madison, WI. Jacquemart, A. L., D. Champluvier, and J. De Sloover. 2003. A test of mowing and soilremoval restoration techniques in we t heaths of the Hig h Ardenne, Belgium. Wetlands 23:376. James, R. T., K. ODell, and V. H. Sm ith. 1994. Water quality trends in Lake Tohopekaliga, Florida, USA: responses to watershed management. Journal of the American Water Resources Association 30:531 536. Jeppeson, E., J. P. Jensen, P. Krist ensen, M. Sondergaard, E. Mortensen, O. Sortkajaer, and K. Olrik. 1990. Fish manipulation as a lake restoration tool in shallow, eutrophic, temperate lakes. 2: Threshold levels, long-term stability, and conclusions. Hydr obiologia 200/201:219 227. Johnson, K. G, M. S. Allen, and K. E. Havens. 2007. A review of littoral vegetation, fisheries, and wildlife responses to hy drologic variation at Lake Okeechobee. Wetlands 27:110 126. Johnson, L. C., and J. R. Matchett. 2001. Fire and gr azing regulate belowground processes in tallgrass prairie. Ecology 82:3377 3389. Keddy, P. A. 1983. Shoreline vegetation in Ax e Lake, Ontario: effects of exposure on zonation patterns. Ecology 64:331 344. Keddy, P. A., and L. H. Fras er. 2000. Four general princi ples for the management and conservation of wetlands in large lakes: the role of water levels, nutrients, competitive hierarchies and centrifugal organizati on. Lakes and Reservoirs: Research and Management 5:177 185. Kershaw, K. A., and J. H. H. Looney. 1985. Quantitative and dynamic plant ecology. Edward Arnold, Baltimore, MD. Killgore, K. J., R. P. Morgan II, and N. B. Rybicki. 1989. Distribution and abundance of fishes associated with submersed aquatic plants in the Potomac River. North American Journal of Fi sheries Management 9:101 111.
128 Kimber, A., C. E. Korschgen, and A. G. van der Valk. 1995. The distribution of Vallisneria americana seeds and seedling light requirements in the upper Mississippi River. Canadian Journal of Botany 73:1966 1973. Lachavanne, J. B. 1985. The influence of accelerated eutrophication on the macrophytes of Swiss Lakes: abundance and distribution. Verhandlungen Internationale Vereinigung fur Theor etische und Angew andte Limnologie 22:2950 2955. Landres, P. B., P. Morgan, and F. J. Swanson. 1999. Overview of the use of natural variability concepts in managi ng ecological systems. Ec ological Applications 9:1176 1188. Langeland, K. A. 1996. Hydrilla verticillata (L.F.) Royle (Hydrocharitaceae), "The Perfect Aquatic Weed". Castanea 61:293 304. Li, S., S. R. Pezeshki, and S. Goodwin. 2004. Effects of soil moisture regime on photosynthesis and growth of cattail ( Typha latifolia ). Acta Oecologica 25:17 22. Light, S. S., and J. W. Dineen. 1994. Water control in the Everglades: a historical perspective. Pages 47 84 in S. M. Davis and J. C. O gden, editors. Everglades: the ecosystem and its restoration. St. Lucie, Boca Raton, FL. Lindig-Cisneros, R., J. De smond, K. E. Boyer, and J. B. Zedler. 2003. Wetland restoration thresholds: can a degradati on transition be reversed with increased effort? Ecological Applications 13:193 205. Little, E. C. S. 1968. The control of water weeds. Weed Research 8:79 105. Mallison, C. T., R. K. Stocker, and C. E. Cichra. 2001. Physi cal and vegetative characteristics of floating islands. Journal of Aquatic Plant Management 39:107 111. McAtee, W. L. 1939. Wildfowl food plant s. Their value, propagation and management. Collegiate Press, Ames, Iowa. McCune, B., and J. B. Grace. 2002. Analysi s of ecological communities. MjM Software Design, Gleneden Beach, OR. McCune, B., and M. J. Mefford 1999. Multivariate analysis of ecological data, version 4.20. MjM Software Desi gn, Gleneden Beach, OR. Milleson, J. F. 1987. Vegetation changes in the Lake Okeechobee littoral zone: 1972 to 1982. Technical Publication. South Florida Water Management District, West Palm Beach, FL. Mitsch, W. J., X. Wu, R. W. Nairn, P. E. Weihe, N. Wang, R. Deal, and C. E. Boucher. 1998. Creating and restoring wetlands. Bioscience 48:1019 1030
129 Mitsch, W. J., and J. G. Gosselink. 1993. Wetlands. Second edition. John Wiley & Sons, New York, NY. Mitsch, W. J., and R. F. W ilson. 1996. Improving the succe ss of wetland creation and restoration with know-how, time, and self-d esign. Ecological Applications 6:77 83. Moss, B., J. Madgwick, and G. Phillips. 1996. A guide to the restoration of nutrientenriched shallow lakes. Borads Authority, Norwich, UK. Moyer, E. J., M. W. Hulon, R. S. Butler, and R. W. Hujik. 1989. Kissimmee chain of lakes studies: study 1. Lake Tohopekali ga investigations. Florida Game and Freshwater Fish Commission, Tallahassee, FL. Moyer, E. J., M. W. Hulon, and J. J. Sweatman. 1995. Fishery responses to habitat restoration in Lake Tohopeka liga, Florida. North Amer ican Journal of Fisheries Management 15:591 595. National Research Council. 1992. Restor ation of aquatic ecosystems: science, technology, and public policy. National Academy Press. National Research Council. 1996. Weltands characteristics and boundaries. National Academy Press. Niering, W. A. 1987. Vegetation dynamics (su ccession and climax) in relation to plant community management. Conservation Biology 1:287 295. Parker, V. T. 1997. The scale of successi onal models and restoration objectives. Restoration Ecology 5:301 306. Pearsall, W. H. 1920. The aquat ic vegetation of the English Lakes. Journal of Ecology 8:164 201. Peet, R. K. 1974. The measurem ent of species diversity. Annual Review of Ecology and Systematics 5:285 307. Poiani, K. A., and W. C. Johnson. 1989. Effect of hydroperiod on seed-bank composition in semipermanent prairie wetlands.Canadian Journal of Botany 67:856 864. Remetrix LLC. 2000. Bathymetric map of La ke Tohopekaliga. Remetrix LLC, Carmel, IN. Sculthorpe, C. D. 1967. The biology of aquatic vascular plants. St. Martin's Press, New York. Schardt, J. D. 1997. Florida aquatic plant survey 1994. Technical Report. Florida Department of Environmental Protection, Tallahassee, FL.
130 Scheffer, M., S. H. Hosper, M-L. Meijer B. Moss, and E. Jeppes en. 1993. Alternative equilibria in shallow lakes. Trends in Ecology and Evolution 8:275 279. Seabloom, E. W., A. G. van der Valk, and K. A. Moloney. 1998. The role of water depth and soil temperature in determining init ial composition of prairie wetland coenoclines. Plant Ecology 138:203 216. Seabloom, E. W., K. A. Moloney, and A. G. van der Valk. 2001. Constraints on the establishment of plants along a fluc tuating water-depth gradient. Ecology 52:2216 2232. Seastedt, T. R., R. J. Hobbs, and K. N. Suding. 2008. Management of novel ecosystems: are novel approaches r equired? Frontiers in Ecology and Environment 6:547 553. Seddon, B. 1972. Aquatic macrophytes as lim nological indicators. Freshwater Biology 2:107 130. Segal, S. 1971. Principles on struct ure, zonation and succession of aquatic macrophytes. Hydrobiologia 12:89 95. Sincock, J. L., J. A. Powell, R. K. Hyde, and H. E. Wallace. 1957. The relationship of the wintering waterfowl populations of the Kissi mmee River Valley to the hydrology, topography, distribution of the vegetat ion and the proposed hydrological regulations. Technical Report. Florida Game and Freshwater Fish Commission, Tallahassee, FL. Smith, D. H., R. M. Smart, and C. G. H anlon. 2004. Influence of water level on torpedograss establishment in Lake Okeechobee, Florida. Lake and Reservoir Management 20:1 13. Sorrell, B. K., H. Brix, H-H Schier up, and B. Lorenzen. 1997. Die-back of Phragmites australis : influence on the distribution and ra te of sediment methanogenesis. Biogeochemistry 36:173 188. Squires, L., and A. G. van der Valk. 1992. Wa ter-depth tolerances of the Delta Marsh, Manitoba. Canadian Jour nal of Botany 70:1860 1867. Stephens, S. L., and L. W. Ruth 2005. Federal forest-fire po licy in the United States. Ecological Applications 15:532 542. Suding, K. N., K. L. Gross, and G. R. Houseman. 2004. Alte rnative states and positive feedbacks in restoration ecology. Trends in Ecology and Evolution 19:46 53. Surber, E. W. 1932. Controllin g vegetation in fish ponds with sodium arsenite. U.S. Dept. of commerce, Investigational Report No. 2:1 39. Tarver, D. P. 1979. Torpedo grass ( Panicum repens L.). Aquatics 1:5 6.
131 Tobe, J. D., K. C. Burks, R. W. Cantrell, M. A. Garland, M. E. Sweele y, D. W. Hall, P. Wallace, G. Anglin, G. Nelson, J. R. Cooper, D. Bickner, K. Gilber t, N. Aymond, K. Greenwood, and N. Raymond. 1998. Florida wetland pl ants: an identification manual. Florida Department of Envir onmental Protection. Rose Printing, Tallahassee, FL. Tilman, D. 1997. Community invasibility, recruitment limitation, and grassland biodiversity. Ecology 78:81 92. Tugend, K. I., and M. S. All en. 2004. Changes in the plant and fish communities in enhanced littoral areas of Lake Kissimm ee, Florida, following a habitat enhancement. Lake and Reservoir Management 20:54 64. Turner, M. G., W. L. Baker, C. J. Peterson, and R. K. Peet. 1998. Factors influencing succession: lessons from large, infr equent natural disturbances. Ecosystems 1:511 523. Turner, M. G., W. H. Romme, and D. B. Tinker. 2003. Surprises and lessons from the 1988 Yellowstone fires. Frontiers in Ecology and Environment 1:351 358. U.S. Army Corps of Engine ers. 1956. Central and southern Florida project for flood control and other purposes: part 2. Kissimmee River ba sin and related areas. General Design Memorandum, Kissimmee Riv er Basin. Office of the District Engineer, Jacksonville, FL. van der Nat, F-J. W. A., and J. J. Middelburg. 1998. Effects of two common macrophytes on methane dynamics in freshw ater sediments. Biogeochemistry 43:79 104. van de Koppel, J., M. Rietkerk, and F. J. We issing. 1997. Catastro phic vegetation shifts and soil degradation in terrestrial graz ing systems. Trends in Ecology and Evolution 12:352 356. van der Valk, A. G. 1981. Succession in wetlands: a Gleasonian approach. Ecology 67:688 696. van der Valk, A. G. 1998. Succession theory and restoration of wetland vegetation. Pages 657-667 in A. J. McComb and J. Davis, editors. Wetlands for the Future. Gleneagles Publishing, Glen Osmond, SA, Australia van der Valk, A. G., and C. H. Welling. 1988. The development of zonation in freshwater wetlands: an experimental approach. Pages 145 158 in H. J. During, M. J. Werger, and J. H. Willems, editors. Diver sity and pattern in plant communities. Academic, The Hague, Netherlands. van Donk, E., and R. D. Gulati. 1995. Transiti on of a lake to turbid state six years after biomanipulation: mechanism s and pathways. Water Science and Technology 32:197 206.
132 Vayssieres, M. P., R. E. Plant, and B. H. Allen-Diaz. 2000. Classification trees: an alternative non-parametric appr oach for predicting species distributions. Journal of Vegetation Science 11:679 694. Walker, B. H., and R. T. Coupland. 1968. An analysis of vegetation-environment relationships in Saskatchewan sloughs. Canadian Journal of Botany 46:509 522. Wegener, W. L., and V. P. W illiams. 1974. Extreme lake drawdown, a working fish management technique. Technical Memor andum. Florida Game and Freshwater Fish Commission, Tallahassee, FL. Wegener, W. L., V. P. Williams, and D. Ho lcomb. 1973. Water level manipulation. Annual Progress Report. Federal Aid in Fish Restoration. Florida Game and Freshwater Fish Commission, Tallahassee, FL. Wetzel, R. G. 1975. Limnology. Saunders College Publishing, Philadelphia, PA. Whisenant, S. G. 1999. Repai ring damaged wildlands: a proc ess-oriented, landscapescale approach. Cambridge University Press, Cambridge, MA. Wilcox, B. A., and J. E. Meeker. 1991. Disturbance effe cts on aquatic vegetation in regulated and unregulated lakes in nort hern Minnesota. Canadian Journal of Botany 69:1542 1551. Williams, J. W., S. T. Jackson, and J. E. Kutzbach. 2007. Project ed distributions of novel and disappearing climates by 2100 AD. Proceedings of the National Academy of Sciences 104:5738 5742. Williams, V. P. 2001. Effects of point-source removal on lake water quality: a case history of Lake Tohopekaliga, Florida. Journal of Lake and Reservoir Management 17:315 329. Wilson, S. D., and P. A. Keddy. 1988. Spec ies richness, survivorship and biomass accumulation along an environm ental gradient. Oikos 53:375 380. Wisheu, I. C., and P. A. Keddy 1992. Competition and centri fugal organization of plant communities: theory and tests. Jour nal of Vegetation Science 3:147 156. Zedler, J. B. 2000. Progress in wetland rest oration ecology. Trends in Ecology and Evolution 15:402 407. Zweig, C. L., and W. M. Kitchens. In press The Semiglades: the co llision of restoration, social values, and the ecosystem concept. Restoration Ecology
BIOGRAPHICAL SKETCH Zach Welch began his college education at the University of Florida in 1994, coming from the small-town high school of D unnellon, FL. Like most, he struggled to find an interest in the first year or two of undergraduate coursework, but was immediately hooked upon taking an introductory wetland ecology class, taught by Dr. Peter Frederick. He earned a Bachelors degree in 1999 majoring in Wildlife Ecology and Conservation, and then began working for Dr Kitchens at the Florida Cooperative Fish and Wildlife Research Unit. He started as a research assistant in the summer of 2000, ev entually earning his Masters degree in 2004 in Wildlife Ecol ogy and Conservation. Having thoroughly enjoyed his experience with Dr Kitchens and the Cooperat ive Research Unit, he eagerly started his PhD progr am with the same advisor, this time majoring in Interdisciplinary Ecology to allow greater emphasis on wetland studies. Throughout his graduate career, wetland ecology and vegetat ion modeling were among his favorite study topics, and he earned his Doctoral deg ree in the fall of 2009 after a long and gratifying experience.