|Table of Contents|
Table of Contents
Chapter 1. Introduction
Chapter 2. Cogongrass invasion displaces sandhill plants and animals
Chapter 3. Cogongrass invasion changes sandhill fire regime
Chapter 4. Cogongrass invasion reduces sandhill seedling recruitment
Chapter 5. Cogongrass invasion limits resources available to sandhill seedlings
Chapter 6. Summary and conclusions
List of references
ECOLOGICAL CONSEQUENCES OF IMPERA TA CYLINDRICA (COGONGRASS)
INVASION IN FLORIDA SANDHILL
CAROL L. LIPPINCOTT
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1997
Dedicated to the gopher tortoise crushed by my sand-skidding truck.
I am sincerely grateful to my advisor, Jack Putz, for always making time to provide prudent and helpful advice and insightful editing. Jack, with his broad knowledge of ecology and infectious enthusiasm for science, encouraged me to expand my thinking about biological invasions. Doria Gordon was a reliable source of advice and a thorough reviewer. I thank Buzz Holling for his unwavering support and encouragement as I ventured into the theory of ecosystem resilience. Stimulating conversations with George Tanner bolstered my confidence and motivated me to work even harder toward conservation and restoration of natural areas. From Donn Shilling I learned to value contributions from the agricultural disciplines toward management of invasive species.
I am grateful to several members of the UF Agronomy Department: Jeff Ray for useful discussions of water limitation experiments, Grady Miller for use of root scanning equipment, Ken Boote and Jean Thomas for use of a plamimeter, Sandra McDonald for sharing her data on cogongrass biology, and Jim Gafffhey for providing his data on rhizome carbohydrates. I thank Kathryn Williams and Russ Pierce of the UF Chemistry Department for use of a bomb calorimeter and Henry Gholz and Ken Clark in the UF School of Forest Resources and Conservation for use of a light ceptometer.
Friends and fellow graduate students offered their valuable time to help with field work: Seth Bigelow, Mary Carrington, Scot Duncan, Mark Edmiston, Jill Fisher, Jeff
Gerwing, Ankila Hiremath, Debbie Kennard, Becky Ostertag, Denise Sauerbrey, and Ann Williams. Staff of the Withlacoochee State Forest, notably Jon Blanchard and Andrea Crisman, and of Ocala National Forest were enthusiastic cooperators in locating field plots and performing prescribed bums. At decisive times I benefited from the wise and encouraging words of Walter Judd, Jack Fisher, Beree Darby, and Bill Stem. I have never worked with anyone as consistently pleasant and dependable as Paula Rowe, the Botany Department secretary.
I am grateful for the support of my family, particularly my mother Margaret,
brother Woody, and sister-in-law Shirley. I would not have attempted and completed this doctoral program without the cheerleading of dear friends, especially Stinger Guala, Kevin McNally, Martha Hoover, Susan Jacobson, Ankila Hiremath, Peggy Olwell, and Mark Edmiston.
A Graduate Research Fellowship from the National Science Foundation allowed me to focus completely on this dissertation work. Additional funding was provided by the University of Florida via a College of Liberal Arts and Sciences Pre-doctoral Fellowship, a Grinter Fellowship, and travel grants form the Graduate Student Council and the College of Liberal Arts and Sciences.
TABLE OF CONTENTS
Biological Invasions and Ecological Resilience .....................................1
Global Grass Invasions.....................6
Cogongrass Invasion and Sandhill Resilience....................................... 9
Florida Sandhill Ecosystems.......................................................... 12
Cogongrass Biology and Ecology ..........................................._....... 15
Cogongrass Invasion in Florida Sandhill .............................. 17
Research Hypotheses.............................................. ................... 19
2 COGONGRASS INVASION DISPLACES SANDHLLL PLANTS AND
Cogongrass Invasion ................................ ......... 21
Displacement of Sandbill. Vegetation ........................................ 23
Displacement of Sandhill Animals ............................... 24
Cogongrass Invasion ................................ ......... 29
Displacement of Sandhill Vegetation ........................................ 30
Displacement of Sandhill. Animals ............................... 31
R esu lts . . . . . .. . .. . . . .. .. . . . .. . .. .. . . . .. . . ... 32
Cogongrass Invasion ................................ ......... 32
Displacement of Sandhill Vegetation ........................................ 36
Displacement of Sandhill Animals............................................ 38
Cogongrass Invasion ......................................... 40
Displacement of Sandbill Vegetation ........................................ 43
Displacement of Sandhill. Animals ............................... 44
3 COGONGRASS INVASION CHANGES SANDHILL FIRE REGIME
Fire in Ecosystems .................................................................... 48
Sandhill Fire Regimne ................................................................. 51
4 COGONGRASS INVASION REDUCES SANDHJLL SEEDLING
Seedling Recruitment ................................................................. 66
Seedling Recruitment in Sandhill. .................................................. 68
Seedlings ........................................................................ 72
Seedlings ........................................................................ 77
5 COGONGRASS INVASION LIMITS RESOURCES AVAILABLE TO
Resource Availability................................................................. 90
Invasive, Plants and Resource Availability ......................................... 92
Florida Sandbill Resources 94
Resource Limitations in Cogongrass................................................ 95
Resource Availability .................................... .... 97
Resource Limitations 100
Seedling Herbivory ............................................................ 103
R esults ...........................................103
Resource Availability .......................................................... 103
Resource Limitations 110
Seedling Herbivory ............................................................ 118
D iscussion ......... .. ............................................. .................................... 118
R esource A vailability ................................................................................ 118
Resource Limitations 125
Cogongrass Allelopathy Review .............................................................. 126
Conclusions .............................................. 127
6 SUMMARY AND CONCLUSIONS
Cogongrass Invasion and Sandhill Resilience ................................................... 129
Cogongrass Invasion Displaces Sandhill Plants and Animals ................ 130
Cogongrass Invasion Changes Sandhill Fire Regime. ............................. 132
Cogongrass Invasion Reduces Sandhill Seedling Recruitment .............. 132
Cogongrass Invasion Limits Resources Available
to Sandhill Seedlings, ....................................................................... 134
Functional Equivalence and Florida Sandhill Resilience .................................. 135
Sandhill M anagement Implications ................................................................... 137
Cogongrass Invasion in Southeastern Ecosystems .......................................... 140
Grass Invasions in Florida Ecosystems ............................................................. 141
Functional Groups as Predictors of Ecosystem Change. .................................. 142
LIST OF REFERENCES 145
BIOGRAPHICAL SKETCH ........................................165
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
ECOLOGICAL CONSEQUENCES OF IMPERATA CYLINDRICA (COGONGRASS) INVASION IN FLORIDA SANDHILL
Carol L. Lippincott
Chairman: Francis E. Putz
Major Department: Botany
I evaluated the functional equivalence of a nonindigenous invader, Imperata
cylindrica (cogongrass), and indigenous grasses in Florida sandhill, a savanna ecosystem. I compared 33 traits of cogongrass and sandhill grasses within key areas of sandhill ecosystem function. I then predicted the effect of cogongrass invasion on the ecological resilience and persistence of Florida sandhill.
Rhizomatous spread of cogongrass was fastest in frequently-burned and clear-cut sandhill. Cogongrass displaced most sandhill vegetation except large trees. Burrowing beetles were displaced by cogongrass but frequency of pocket gopher mounds was not changed. Cogongrass was a lower-quality forage than sandhill vegetation.
Cogongrass fires were higher, hotter, more continuous horizontally, and killed more juvenile longleaf pines than sandhill fires. Fuel accumulated more rapidly in cogongrass than in sandhill, so fire could ignite more frequently.
Cogongrass reduced recruitment of planted seeds and seedlings of 7 sandhill
species representing various life forms. The only survivors from planted seeds were of a large-seeded palm. In cogongrass, clonal species planted as seedlings had more survivors and a N-fixing clonal shrub grew better than other types of seedlings. Fewer planted pine seedlings survived in cogongrass, and survivors were shorter than in sandhill. Burning before planting was equally important to seedling recruitment in cogongrass and sandhill.
Relative to sandhill, cogongrass litter reduced light more at ground level.
Cogongrass' higher leaf area, and greater rhizome and root mass and root length density reduced soil water more than in sandhill. Soil nutrient availability was apparently not changed by cogongrass. Increased growth of seedlings in exclosures was evidence of light and water limitations in cogongrass. Seedling survival depended strongly on water availability, while growth depended on both water and nutrients.
Tall rhizomatous cogongrass is not functionally-equivalent to the short, caespitose grasses indigenous to sandhill. Cogongrass represents a new functional group that can exceed sandhill resilience through changes in vegetation structure, soil processes, animal populations, fire regime, seedling recruitment, and resource availability. Cogongrass has the ability to convert Florida sandhill to treeless species-poor grassland. Comparisons of functional equivalence between invaders and key indigenous species may have utility in predicting ecosystem change caused by nonindigenous species.
This dissertation examines the ecological consequences of invasion by Imperata cylindrica (cogongrass) in Florida sandhill ecosystems. Specifically I ask if cogongrass, a nonindigenous species, can significantly alter sandhill ecosystem function in ways that exceed resilience of the system, resulting in a shift from pine forest to grassland dominated by cogongrass. In this chapter I begin by discussing biological invasions of nonindigenous species and their affects on ecosystem structure and function, emphasizing invasive plants. I then incorporate these ideas into current concepts of ecological resilience as a means to predict and interpret the effects of invasive species on ecosystem persistence. I then review the global role of invasive grasses in altering ecosystem function. Finally, I focus on the dynamics of cogongrass invasion in Florida sandhill ecosystems and present this as a test case for determining whether an invasive species can erode ecosystem resilience to the point of causing a shift to a new type of ecosystem.
Biological Invasions and Ecological Resilience
Biological invasions of nonindigenous species are an under-emphasized component of global environmental change, comparable to widely recognized phenomena such as land-use change and elevated atmospheric carbon dioxide (Vitousek et al. 1996, Huenneke 1997). As agents of global change, invasive nonindigenous species are reducing the earth's
biological diversity through the extirpation of genetically distinct populations and extinction of species (OTA 1993, Vitousek 1997).
Carried by humans across biogeographic barriers to regions where they never
existed before, many plants and animals have established and spread outside of cultivation, with unintended, unexpected, and undesirable ecological consequences. Once established in natural ecosystems, nonindigenous species can result in dramatic reductions in indigenous species diversity, species extinctions, and profound changes in ecosystem structure (physiognomy) and function (productivity, nutrient cycles, trophic patterns, water use, etc.) (Elton 1958, Mooney and Drake 1986, Drake et al. 1989, Cronk and Fuller 1995, Pysek et al. 1995, Williamson 1996, Luken and Thieret 1997). As a result, species-rich ecosystems are being fundamentally changed at unprecedented rates by a relatively small subset of cosmopolitan invaders. Since ecosystems provide resources such as clean water, food, and recreation opportunities that are essential to human welfare, invasive nonindigenous species have important economic and social impacts in addition to ecological effects (Folke et al. 1996, Daily 1997).
There are numerous well-documented cases of environmental change by invasive nonindigenous animals. For example, feral ungulates are some of the most damaging introductions in national parks and preserves in the United States, where they disturb soil and facilitate dispersal of invading plants (Coblentz 1990). In the 13 years since Dreissena polymorpha (zebra mussel) was inadvertently carried in ballast water of ships from Eurasia to the Great Lakes of North America, it has rapidly covered the bottoms of rivers and lakes. Zebra mussels displace indigenous clams, and filter water so efficiently that phytoplankton and consequently zooplankton and pelagic fish decline sharply (Caraco et
al. in press). Introduction of Lates niloticus (Nile perch) into Lake Victoria in east Africa in the late 1950s resulted in the extinction of 2/3 of the 300 endemic cichlid fish within 40 years, the largest documented extinction directly due to a nonindigenous species (Witte et al. 1995).
There are also well-documented examples of invasive nonindigenous plant species that fundamentally alter ecosystems (Ramakrishnan and Vitousek 1989, Vitousek 1990, D'Antonio and Vitousek 1992, Vitousek et al. 1996). Some affect biotic processes, displacing indigenous species, altering species diversity and dominance (including keystone species complexes), and changing productivity, decomposition, and nutrient cycling. Others have been shown to change abiotic processes such as hydrology, soil chemistry, geomorphology, and disturbance regimes. For example, Lythrum salicaria (purple loosestrife) from Eurasia has completely displaced wetland plant communities across temperate North America (Thompson et al. 1987). Sapium sebiferum (Chinese tallow) woodlands dominated former coastal prairie in Texas, habitat of the endangered Texas prairie chicken (Bruce et al. 1995, Bruce et al. 1997). Melaleuca quinquenervia (Australian cajeput), Schinus terebinthifolius (Brazilian pepper), and Lygodium microphyllum (climbing fern) dominated large portions of the Florida Everglades (Schmitz and Brown 1994, Simberloffet al. 1997), and a majority of the 31 most invasive species in Florida show traits capable of modifying ecosystem properties (Gordon (in press)). Two species of Tamarix (saltcedar) displaced indigenous plants, increased soil salinity, and lowered the water table in arid riparian habitats of the southwestern United States (Loope et al. 1988). Invasion by nitrogen-fixing Myricafaya (firetree) on recently formed volcanic soils in Hawaii caused a rapid 4-fold increase in the amount of biologically available
nitrogen in the soil, facilitating invasion of other nomindigenous plants (Vitousek and Walker 1989). Invasion of Mesembryanthemum crystallinum (ice plant) in California coastal grasslands redistributed salt from throughout the rooting zone onto the soil surface, interfering with growth of other species and increasing soil erosion (Vivrette and Muller 1977). Finally, nonindigenous grass invasions in Hawaii increased the frequency and size of woodland fires, converting indigenous forest communities to swards of nonindigenous grass (Smith and Tunison 1992).
These cases show that invasion by nonindigenous plants and animals can
fundamentally alter ecosystem structure and function, eventually leading to loss of existing ecosystems with replacement by simpler ecosystems of uncertain composition. In attempting to understand how I or a few invasive species can so drastically alter diverse ecosystems, I find it helpful to invoke the concept of ecological resilience. Ecological resilience, as used here, is the magnitude of disturbance that can be absorbed before the ecosystem's controlling processes change to the extent that the structure of the system is essentially changed (Walker et al. 1969, Holling 1973). This definition of resilience assumes that an ecosystem can exist in any of several alternative stable states or "stability domains." Ecological resilience focuses on the ability of a set of structures and processes to persist, and examines conditions far from any equilibrium, in which rare and unpredictable disturbances can flip a system into another stability domain (Holling 1986, 1992, Holling and Meffe 1996). This differs from the more traditional concept, termed engineering resilience (Holling 1994), which measures resistance to disturbance and speed of return to the equilibrium steady state, and assumes that a system will constantly
remain near I equilibrium steady state (Pimm 1984, ONeill et al. 1986, Tilman and Downing 1994).
Ecosystems are primarily structured by interactions between a key subset of biotic and abiotic variables that entrain a diversity of other species that are affected by, but do not usually affect, ecosystem function. Resilience to disturbance is determined by the strength of interactions between these key control variables and by the diversity of overlapping influences within these variables (Holling et al. 1995, Folke et al. 1996, Holding and Meffe 1996).
In deciphering the role of invasive species in exceeding ecosystem resilience, i.e., shifting from I stability domain to another, it is useful to consider the "drivers and passengers" hypothesis of ecological function (Walker 1992, 1995). This hypothesis states that ecological function primarily resides in driver species that act in ecologically similar ways within functional groups (Solbrig 1993, Huston 1994, Solbrig et al. 1996) to strongly influence ecosystem processes. Drivers are the species that control the future of the ecosystem while passenger species have minor ecological roles (although in certain circumstances some passengers may become drivers). Ecological resilience of an ecosystem is greater when there is a diversity of driver species within each functional group. These driver species are functionally equivalent and have different but overlapping responses to disturbances. This overlap allows the ecosystem to absorb the impacts of unusual disturbances and to persist, i.e., the stability domain and resilience increase. Conversely, reduction in the number of driver species can result in reduced ability of the system to absorb novel disturbances, i.e., the stability domain and resilience decrease (Holling 1992, Walker 1995). If novel disturbances cannot be absorbed because of low
diversity of driver species, then the ecosystem shifts to a new stability domain characterized by a different set of processes, structures, and species.
Incorporating the concept of ecological resilience and the "drivers and passengers" hypothesis of ecosystem function, I propose that addition of new species can strongly influence ecosystem function in sudden and unexpected ways that exceed the resilience of the system. I suggest that this is the case with addition of nonindigenous species, where the most damaging invaders are those that represent a new functional group. These functionally dissimilar invasive species could in effect be novel disturbances, driving the system to a different stability domain. I examined this proposition in my study of invasion of a nonindigenous grass in a species-rich resilient ecosystem already dominated by grasses. First, I will review the importance of grass invasion worldwide.
Global Grass Invasions
Numerous grass species have been carried around the world as forage crops or for soil erosion control (Hartley and Williams 1956), and many of the traits that agronomists have traditionally used to select promising forages also identify "ideal weeds" (Baker 1965): they regrow rapidly after burning, grazing, and mowing; can resprout from rhizome fragments after plowing; and are highly competitive in a variety of environmental conditions (Chapman 1996). As a result, 12% of the world's total agricultural weed species are grasses, and the family Poaceae is a close second to the Asteraceae in number of genera (166) and species (753) that are weeds (Heywood 1989). Indeed, 10 of the 18 most noxious agricultural weeds in the world are grasses (Holm et a]. 1977), the majority of these (21 out of 3 1) in the subfamily Panicoideae, which includes the genus Imperata (Chapman 1996). Afican and Asian grasses, which have evolved perennating organs near
or below the ground and regrow rapidly after defoliation in response to intense ungulate grazing (Parsons 1972), were widely distributed because of these traits, and are now some of the most globally widespread plant invaders (Hartley and Williams 1956). Seventeen of the 31 principal grass weeds in agricultural areas worldwide are of African origin (Chapman 1996).
In addition to being noxious weeds in agricultural areas, grass invasions in natural areas are now recognized as a worldwide problem (D'Antonio and Vitousek 1992). In a recent global taxonomic analysis of invasive plants (Daehler 1997), grasses were included with those plants that represent the highest risk of becoming natural-area invaders. The following are examples of invasive nonindigenous grasses, grouped by geographic area of invasion.
In South America, several African grasses [Panicum maximum (guinea grass),
Brachiaria mutica (Para grass), Melinis minutiflora (molasses grass), Hyparrhenia rufa (jaragua), Pennisetum clandestinum (Kikuyu grass), and Digitaria decumbens (Pangola grass)] now dominate vast areas of savanna (Baruch 1996). Studies of African grass invasions in South American savannas have demonstrated drastic reductions in species and structural diversity while increasing primary productivity, suggesting impacts to ecosystem processes such as nutrient cycling, hydrology, fire, grazing, and succession (Baruch 1996, Solbrig et al. 1996). For example, molasses grass imported to support intensive ungulate grazing in cleared Venezuelan forest lands subsequently invaded savanna dominated by the indigenous grass Trachypogon plumosus (Baruch and Gomez 1996), particularly as soil nutrient availability increased after years of fire suppression (Bilbao and Medina 1990). Two deep-rooted African grasses, Andropogon gayanus and Brachiaria humidicola,
sequestered carbon at deep layers in the soil profile in South American savannas, with possible consequences for global carbon cycles (Fisher et al. 1994).
Most of the forage grasses successfully introduced into Australia are also African species (Cox et al. 1988), and 20% of the pasture grasses that performed well in Australian agronomic field trials later became established weeds in Kakadu National Park (Lonsdale, personal communication). In Australia, nonindigenous grasses have been shown to inhibit regeneration of indigenous shrubs and trees, alter fire regime, and convert woodlands to grasslands, and are implicated in the decline of rare plant species (Humphries et al. 1991). In the moist tropical zone of Australia, molasses grass has displaced native grasses that colonize forests after cyclone damage, arresting natural succession to forest (Gill et al. 1990).
In the United States, invasion by Bromus tectorum (cheatgrass) is a welldocumented case of a European grass invading and dominating rangeland in the Great Basin region (Mack 1986). This annual species rapidly established in soil disturbed by cattle, and depleted soil water at times of year critical for recruitment of indigenous plant species. In addition, cheatgrass increased both the size and number of fires in the region, converting higher elevation woodlands to grassland (Klemmendson and Smith 1964). In Californiats valley grasslands, grazing facilitated invasion of shallow-rooted Mediterranean annual grasses (Bromus mollis, Bromus diandrus, Avena spp.), displacing grazingintolerant deep-rooted perennial bunchgrasses, increasing soil moisture in lower soil layers, and possibly promoting increases of another nonindigenous invader, Centaurea solstitialis (yellow star thistle), that depletes deep soil moisture reserves (Heady 1977, Rice et al. 1997).
In the seasonal submontane forests in Hawaii, nonindigenous perennial grasses created fire cycles in ecosystems where fire previously did not occur because they represent a new growth form in an ecosystem previously dominated by woody, evergreen species (Mack and D'Antonio 1997). In these forests, where grasses previously constituted about 30% of the understory biomass, grasses now constitute over 90% of the living understory biomass (D'Antonio et al. 1997).
Cogongrass Invasion and Sandhill Resilience
I examined invasion of an Asian grass, Imperata cylindrica (cogongrass), in sandhill ecosystems in the state of Florida in southeastern North America. This nonindigenous grass has been spreading across the Coastal Plain region of the southeastern United States since its introduction in 1911 to 1912. Although it is one of the world's worst weeds in agricultural areas (Holm et al. 1977), its threat to southeastern ecosystems such as sandhill was less certain. Sandhill consists of species-rich pine forests with a grass-dominated understory and extensive soil mixing by burrowing animals. Sandhill grasses, pines, and burrowers (key biotic processes) interact with well-drained soils, seasonal drought, frequent low intensity fires, and occasional hurricane-force winds (key abiotic processes) to structure the sandhill ecosystem. As a result of strong interactions between these biotic and abiotic processes and overlap within functional groups, the sandhill ecosystem is considered resilient to novel disturbance, perhaps including novel grass species.
Fire is a key controlling process in sandhill, important in maintaining its
characteristic structure and function (Christensen 1987). Sandhill fires are primarily fueled by short caespitose grasses and pine needles, the "drivers" in the functional group that
perpetuates fire. Similar plant functional groups have been described for savanna ecosystems (Baruch et al. 1996). In this pyrogenic grass-dominated system, cogongrass might be just an additional driver species, functionally equivalent to sandhill grasses and increasing the stability domain and resilience of the system. If cogongrass were functionally equivalent to sandhill grasses, its spread into sandhill would not significantly change important sandhill processes. Arrival of this new grass would increase sandhill resilience and the system's characteristic structuring processes would continue to function as before. Sandhill would persist, but with a new addition to its herbaceous understory.
If, on the other hand, cogongrass were not equivalent to sandhill grasses, it could in effect be a novel disturbance (Sousa 1984), defined here as a relatively discrete event that disrupts ecosystem, community, or population structure and changes resource availability, substrate conditions, or physical environment parameters (Pickett and White 1985). For example, cogongrass might change plant community structure as it displaces sandhill plants, changes litter (fuel) arrangement and mass and thus alters fire behavior, enriches the soil organic matter layer, alters soil water dynamics, or displaces burrowing animals. These multiple changes would be considered novel disturbances that could erode the resilience of sandhill, resulting in a shift from pine forest to a new stability domain of grassland dominated by cogongrass.
From this perspective, I studied the ecological consequences of cogongrass
invasion in Florida sandhill to test the general hypothesis that cogongrass is functionally equivalent to sandhill grasses. I designed experiments to examine cogongrass' effects on key biological and physical processes in sandhill, presented graphically in a conceptual model in Figure 1-1. If cogongrass is equivalent to sandhill grasses, then I would detect no
.. ... .. ...... .... ... .... .. ...
... .... ....
..... .. . ..
. ... .. . .. ..4.
..... ..... .. .. .....
.... .... ....
..... .... .... .... .... ....
.... .. .... ... ... .... ... ...
..... ..... ..0..
differences in these key processes, and sandhill would likely persist as a pine forest but with a new grass component in its understory. If, however, I detected significant changes to key sandhill processes, then cogongrass should not be considered a functional equivalent of sandhill grasses. In this case, cogongrass would create disturbances that will likely exceed the resilience of this ecosystem, and sandhill would eventually be converted from pine forest to grassland dominated by cogongrass.
Florida Sandhill Ecosystems
Forests of longleaf pine once dominated the southeastern Coastal Plain from
southeastern Virginia to east Texas (Myers and Ewel 1990). Across its range, longleaf pine is found on poorly drained spodic soils in flatwoods ecosystems and on well-drained sandy soils in high pine ecosystems. High pine ecosystems are characterized by an open overstory of pine and a grass-dominated understory interspersed with clonal oaks, on soils ranging from loamy sands underlain by clay (clayhill) to coarse well-drained sands (sandhill). Sandhill soils are low-nutrient entisols (Typic Quartzipsamments) derived from marine sand deposits and lacking profile development (Myers and Ewel 1990). In contrast, clayhill soils are higher-nutrient ultisols with distinct profile development. In Florida, sandhill (Figure 1-2) extends from the middle of the Florida peninsula to the northwestern border of the state (Davis 1967). In north Florida, sandhill gradually grades northward into clayhill. Average annual precipitation across the range of Florida sandhill ranges from 119 to 163 cm/yr, most occurring from June to September (Winsberg 1990).
The vegetation of Florida sandhill is characterized by an overstory of Pinus
palustris (longleaf pine) except in south-central Florida where Pinus elliottii var. densa (south Florida slash pine) dominates. Below the pines is a varying degrees of subcanopy of
Withlacooche State Forest
(adapted from Davis, 1967)
Figure 1-2. Map of distribution of Florida sandhill, including location of the two field study sites for this research project.
fire-tolerant oak such as Quercus laevis (turkey oak), and an herbaceous understory of small-statured herbs and short perennial caespitose grasses such as Aristida stricta (wiregrass), and Sporobolus spp. (pineywoods dropseed). These grass taxa are very similar morphologically and are often confused (Hall 1989), suggesting a strong functional equivalence. Longleaf pine forests in the Southeast often have more than 40 plant species/m2, with many endemic species (Peet and Allard 1993). One-quarter of the rare plant taxa (45 spp.) in the Southeast are located in longleaf pine forests; 35 of these occur only in Florida sandhill ecosystems (Walker 1993).
Burrowing animals, including Geomyspinetis (southeastern pocket gopher), Gopherus polyphemus (gopher tortoise), Peltotrupes youngi (scarab beetle), and Pogonomyrmex badius (eastern harvester ant), are important soil processors that prevent soil profile development in sandhill (discussed in Chapter 2) In addition, gopher tortoise burrows are occupied by over 300 vertebrate and arthropod species (Jackson and Milstrey 1988). While many of these commensal animals also use other habitats, some such as Rana areolata (gopher frog) inhabit only tortoise burrows. Because of the importance of tortoise burrows to commensal animals, Gopheruspolyphemus is considered a keystone species in sandhill ecosystems (Eisenberg 1983).
In addition to burrowing animals, a number of threatened or endangered vertebrate animals live in sandhill: Picoides borealis (red-cockaded woodpecker), which nests only in cavities in mature pines; Aimophila aestivalis (Bachman's sparrow), which nests on the ground in wiregrass clumps; Sciuris niger (southeastern fox squirrel); and Pituophis melanoleucus (pine snake), Peromyscusfloridanus (Florida mouse), and Drymarchon corais couperi (eastern indigo snake), all of which inhabit gopher tortoise burrows. Sixty-
nine percent of the mammals and 36% of the characteristic bird species in longleaf pine forests forage primarily on or near the ground, an indication of the importance of groundcover maintenance by fire (Engstrom 1993, Ware et al. 1993).
Sandhill is also habitat for 4000 to 5000 species of arthropods, with perhaps 10% of these endemic to this ecosystem (Folkerts et al. 1993). Over 70% of the plants in sandhill are insect-pollinated, and herbivores and litter detritivores may have important roles in sandhill ecosystem function. Fire exclusion, silvicultural practices, or nonindigenous species invasion could have major negative effects of arthropod diversity and sandhill function.
Prior to European settlement of the Coastal Plain, longleaf pine forests occupied about 22 million ha, the most extensive forest type in the region (Frost 1993). Today, about 1 million ha of second-growth longleaf pine forest remains in the Southeast, in addition to about 600 ha of old growth forest in small isolated stands (Simberloff 1993). The rest has been converted to cropland, pasture, pine plantation, and urban uses, or changed to hardwood forests due to fire exclusion. In Florida, 385,000 ha of sandhill forest remain. Almost half of this forest is in public ownership, managed for timber extraction, hunting, wildlife management, recreation, or wilderness preserves. Florida sandhill is an important social and economic resource, generating jobs, recreational and hunting opportunities, and revenue from wood products (Landers et al. 1995).
Cogongrass Biology and Ecology
In this study I follow the taxonomic treatment in the most recent monograph of the genus Imperata (Gabel 1982). Imperata cylindrica (L.) Beauv. (family Poaceae, subfamily Panicoideae, tribe Andropogoneae, subtribe Andropogoninae) is a highly variable species
with a wide geographic range, from the western Mediterranean to South Africa, through India, Southeast Asia, and the Pacific Islands to Australia. Although 5 varieties of cogongrass were described (Hubbard et al. 1944), they are considered invalid by current criteria because of the range of variation within each variety. Imperata brasiliensis (Brazilian satintail), indigenous to southern Florida and now found throughout the Southeast (Watson and Dallwitz 1992), is reportedly distinguished from cogongrass by having only 1 stamen per flower instead of 2, although this may be a more variable character than previously thought.
Cogongrass is a perennial grass with C4-type photosynthesis that grows to 1.5 m tall in erect tufts from extensive scaly rhizomes. Tillers are spaced from 1 cm to 1 m apart on rhizomes. Leaves are mostly basal, and do not have auricles. Leaf blades are I to 2 cm wide and up to 1.5 m long, have rough margins, and taper to a narrow base. The upper surface of the leaf is hairy near the base and the lower surface is smooth. Leaf sheaths can be hairy or smooth. Ligules are a fringed membrane about 0.7 to 1.7 mm long. Panicles are cylindrical, spikelike, and silvery, to 3.5 cm wide and 6 to 22 cm long. Glumes are similar, membranous with long silvery hairs especially towards the base. Spikelets are paired, consistently in "long and short" combinations, 3.5 to 4.3 mm long, and have 2 stamens per flower. The oblong caryopsis is about 1 mm long (Gabel 1982, Hall 1983, Watson and Dallwitz 1992).
Cogongrass grows under a broad range of environmental conditions in warm temperate and tropical zones in 73 countries worldwide, including the southeastern Coastal Plain region and peninsular Florida in the United States. It has colonized deserts, sand dunes, grasslands, forests, river margins, and swamps. Cogongrass is one of the
most troublesome invasive plants worldwide (Holm et al. 1977). Because of its ability to invade agricultural areas, cogongrass is listed under the United Statest Federal Noxious Weed Act, 1990 revision. It is also included in the Florida Department of Agriculture and Consumer Service's Noxious Weed List (Florida Statutes, Chapter 5B-57.007, 1993 revision) and on the Florida Exotic Pest Plant Council's 1997 invasive plant list.
Cogongrass reproduces by both rhizome extension and seed spread. Within its indigenous range, cogongrass prefers moist acidic soils and is limited mostly by extreme aridity (Hubbard et al. 1944). First flowering occurs within 1 year of germination. Seeds germinate soon after ripening, have no dormancy, and establish in open disturbed areas. In Southeast Asia, rhizomes mostly develop 10 to 30 cm below the surface and form dense extensive layers that readily produce above-ground tillers after burning, grazing, or mowing (Ayeni and Duke 1985). Studies in Southeast Asia have reported allelopathic effects of cogongrass exudates, discussed in Chapter 5 (Eussen 1977).
Cogongrass Invasion in Florida Sandhill
Documentation of the arrival of cogongrass into the southeastern United States is scarce, but there are reports of 2 separate introductions. In 1911 to 1912, cogongrass was accidentally introduced near the town of Grand Bay (Mobile County), Alabama, as discarded lining of boxes of bare-root orange plants shipped from Japan (Tabor 1952a, 1952b). It was also intentionally introduced as a forage plant from the Philippines and established at the Mississippi Experiment Station in McNeil (Pearl River County) around 1920. A few years later, it was planted at the Florida Experiment Station in Gainesville (Alachua County), and by 1934 it was grown in an experimental planting in Brooksville (Hernando County), Florida. Plantings were also made on the firebreaks of the nearby
Withiacoochee Land Utilization Project. A small test plot of cogongrass was established at Auburn (Lee County), Alabama, at the Agricultural Experiment Station farm in 1943, but this planting was destroyed in 1949 because of the plant's reputation as a pest (Hubbard et al. 1944, Tabor 1949, 1 952b, Dickens 1974).
The Florida Agricultural Experiment Station's policy in the 1930s was to prohibit release of cogongrass propagules due to its potential danger as a pest and because other pasture grasses were considered superior to it. However, in the 1930Os cogongrass was apparently taken without authorization from the Florida Experiment Station in Gainesville and planted in northwest Florida near DeFuniak Springs (Walton County). From there it was carried by cattlemen to central Florida near Ocala (Marion County) and perhaps other portions of the state (Tabor 1949, 1952b).
In the late 1940s and early 1950Os, scientists familiar with the weedy nature of cogongrass in other parts of the world wrote of their concerns about its presence in the United States and Puerto Rico and strongly recommended its eradication from the western hemisphere (Pendleton 1948, Tabor 1949, 1952b). In spite of these early warnings, cogongrass was not controlled and has spread across the southeastern United States and is now found in Florida, Georgia, South Carolina, Alabama, Mississippi, Louisiana, and Texas (Bryson and Carter 1993). Throughout the Southeast, cogongrass is a weed in pastures, plantations, and other managed areas (Patterson et al. 1980). In Florida, cogongrass is documented in 22 Florida counties (Wunderlin et al. 1997) and is considered a major pest on highway roadsides (Willard et al. 1990). Cogongrass also occurs in dense continuous swards covering thousands of hectares in the heavy clay soils of abandoned phosphate mines in Polk County in central Florida (Coile and Shilling 1993).
In addition to roadsides and other chronically disturbed areas, cogongrass has invaded natural ecosystems in the Southeast such as sandhill, flatwoods, and hammock edges. In 1994 1 visited and/or spoke with managers of most of the public lands in Florida containing longleaf pine-dominated sandhill. I found that cogongrass occurred, to various degrees, in most of the remaining publicly owned sandhill ecosystems in the state, apparently introduced both by rhizomes and seed. On these public lands, cogongrass control efforts range from aggressive chemical eradication, as in Ocala National Forest and Withlacoochee State Forest, to no control due to lack of funds or lack of awareness of its damaging potential.
In summary, this study was designed to examine the ecological consequences of cogongrass invasion in Florida sandhill ecosystems. I tested the effects of cogongrass on key ecological processes in sandhill to determine if cogongrass is functionally equivalent to sandhill grasses. With these results I predicted the effect of cogongrass invasion on resilience of sandhill ecosystems. Specifically, I designed experiments to examine the following 4 hypotheses:
* Cogongrass invasion displaces indigenous sandhill plants and animals.
* Cogongrass invasion changes fire regime in sandhill. Cogongrass invasion reduces sandhill seedling recruitment. Cogongrass invasion limits resources available to sandhill seedlings. The results of experiments carried out to test these hypotheses are presented in the 4 subsequent chapters. In the final chapter, I summarize these findings to predict the effect of cogongrass invasion on sandhill resilience, and discuss the utility of this approach in predicting ecosystem change caused by nonindigenous species invasion.
COGONGRASS INVASION DISPLACES SANDHILL PLANTS AND ANIMALS Introduction
The purpose of the studies presented in this chapter is to ascertain whether
cogongrass is a functional equivalent to sandhill grasses in terms of the biotic processes of invasion and consequent displacement of sandhill plants and animals. First I measured rates of vegetative spread (invasion) of cogongrass, as influenced by various sandhill management practices. These data were then incorporated into mathematical models of cogongrass spread over 20 yr. Next I sought evidence of displacement of sandhill vegetation by measuring cover of different strata of sandhill vegetation in cogongrass of various coverages, and by examining frequency of large trees in dense cogongrass. Finally, I looked for evidence of displacement of burrowing animals by comparing frequency of their mounds, and of grazing/browsing animals by comparing nutritional quality of shoots and roots of cogongrass and of sandhill plants.
If these studies indicate that cogongrass significantly alters these biotic processes in Florida sandhill, this supports the hypothesis that cogongrass is not functionally equivalent to sandhill grasses. In this case, these data would provide evidence that cogongrass represents a new functional group in sandhill ecosystems and that this invader has the potential to exceed sandhill's resilience, shifting this ecosystem to a new stability domain, of grassland dominated by cogongrass.
Imperata cylindrica (cogongrass) spreads by wind-dispersed seed and lateral
extension of rhizomes, allowing cogongrass swards to cover large areas, as it has across millions of hectares in Southeast Asia (Holm et al. 1977). In its indigenous range, spread and dominance of cogongrass is largely due to human clearing of tropical rain forest with subsequent burning (Eussen and Wirjahardja 1973). Throughout the Old World tropics, dominance of cogongrass is enhanced by frequent burning (Brook 1989). Its introduction into the United States first occurred in 1911 1912 (Hubbard et al. 1944, Tabor 1949), and cogongrass has since spread across the southeastern Coastal Plain (see Chapter 1). Cogongrass is now legally recognized as a noxious weed by the US Federal Noxious Weed Act of 1990, and its intentional cultivation and distribution is prohibited in the United States. Cogongrass continues, however, to spread by wind dispersal of seed and unintentional human transport of rhizomes in fill soil and on farm and road maintenance equipment (Dickens 1974, Willard et al. 1990).
While long-distance dispersal of cogongrass in the Southeast is primarily due to inadvertent rhizome transport, seed can contribute to local spread (Wilcut et al. 1988). A recent study (Shilling et al. 1997) showed that cogongrass spikelets were dispersed by wind within 110 m of the inflorescence and that cogongrass seed viability was variable. Seed from 9 sites in central Florida showed a high variance between sites, with mean spikelet fill ranging from 0 to 39% and viability of excised caryopses ranging from 0 to 100%. Since cogongrass is mostly self incompatible (Santiago 1965, Gabel 1982), low seed viability may be partly due to long distances between clones of different genotypes.
The identification and distribution of cogongrass genotypes in the Southeast is not known. A recent study, however, examined genetic variation of cogongrass at 10 sites in central and north Florida (Shilling et al. 1997). Random amplified polymorphic DNA analysis (RAPD) techniques were applied to DNA extracted from green leaves. A similarity matrix (for 1 primer) based on the number of shared RAPD bands among the 10 sites showed that no 2 sites had a similarity value higher than 0.47 (identical DNA has a similarity value of 1. 00). These data suggest that these 10 cogongrass populations are genetically different from each other. The same procedure applied to tillers from within populations showed no genetic diversity for a population with no viable seed and high genetic diversity for a population with high seed viability. These data again suggest that low seed viability may be due to long-distances between clones of different genotypes, although variability in environmental conditions also influences seed formation and viability from year to year (Connor 1987, Chapman 1992, 1996).
Cogongrass is abundant in Citrus County, Florida along roadsides and in pastures surrounding Withlacoochee State Forest. By 1934 it was being cultivated at an experiment station in Brooksville (Hubbard et al. 1944) and in the 1930s it was distributed to central Florida farmers and ranchers (Tabor 1949, 1952a). In addition, cogongrass was planted along firebreaks in the "Withiacoochee Land Utilization project" near Brooksville (Tabor 1949). Staff at Withiacoochee State Forest speculate that cogongrass was accidentally introduced into the forest in roadfill from nearby clay pits infested with cogongrass, beginning at least 20 yr ago. This hypothesis is supported by the pattern of cogongrass distribution here, where it occurs in undulating bands, usually less than 35 m wide, along interior forest roads.
To estimate the rate of vegetative spread of cogongrass in Florida sandhill and to understand how forest management practices influence cogongrass spread, I measured vegetative spread in sandhill managed for multiple purposes--pine harvest, wildlife management, and biodiversity protection. Since all of these management goals call for prescribed fire, I measured vegetative spread of cogongrass in fire-managed vs. fireexcluded sandhill. Since pine harvest (i.e., clearcutting and artificial regeneration) results in sudden removal of tree canopy and soil disturbance ranging from passage of heavy extraction equipment to roller-chopping, I also studied the effects of clear-cutting on cogongrass spread. To investigate whether small isolated clones of cogongrass are capable of spreading into diverse intact sandhill vegetation, I transplanted plugs of cogongrass into sandhill.
Since these data measured spread over only 2 years, I used these and other data to construct 4 models of cogongrass spread in sandhill, extrapolated over a 20-year period. I limited the models to this relatively short time period because of the limitations found in spatial spread (diffusion) models when extended beyond initial period of spread due to stochasticity, environmental heterogeneity, and age structure (Lonsdale 1993, Hastings 1996, Higgins and Richardson 1996, Higgins et al. 1996). Displacement of Sandhill Vegetation
Cogongrass is known to establish and spread in open areas such as pastures,
roadsides and pine forests throughout the Southeast (Bryson and Carter 1993). It is not known, however, whether cogongrass invasion in sandhill results in its exclusive dominance and elimination of existing indigenous sandhill vegetation. If, for example, cogongrass only displaces the sandhill grasses and herbs but has no affect on existing
woody plants, then it might be considered a substitute for sandhill grasses if one were managing solely for growth of long-lived palms, shrubs, and trees. (The obvious next question would then be "How does cogongrass affect seedling recruitment of woody species?," discussed in Chapter 4). In this study, I examine the relationship between cogongrass invasion and dominance of existing sandhill vegetation. I hypothesized that as cogongrass cover increases, cover of sandhill vegetation decreases, and that herbaceous sandhill vegetation is displaced before woody species, while large trees are not affected. Displacement of Sandhill Animals
Burrowers. A guild of fossorial animals inhabits the dry sandy soils of sandhill. Soil mixing by burrowing animals is an important soil process in sandhill, retarding formation of distinct soil profiles (Kalisz and Stone 1984). Burrowing animal mounds differ in edaphic characteristics and microclimate from surrounding undisturbed areas, which may have implications for seedling recruitment (Kalisz and Stone 1984). By creating above-ground soil mounds, these burrowing animals may influence plant species composition and plant community structure and function (Hermann 1993).
Three common fossorial animals in southeastern sandhill, Gopheruspolyphemus (gopher tortoise), Geomyspinetus (southeastern pocket gopher), and geotrupine scarab beetles such as Peltotrupesyoungi, burrow to different depths: scarab beetles the deepest (from 3.6 to 5.0 m), gopher tortoises shallower (from 1.8 to 2.8m), and pocket gophers nearest to the soil surface (to 0.6 m). Gopher tortoise mounds, known as "aprons," average 1 m2 or more in diameter and are distinguished by a visible burrow entrance to one side. Scarab beetles and pocket gophers produce smaller round mounds (< 10 cm and 20
to 30 cm diameter, respectively) that cover burrow entrances (Kalisz and Stone 1984, Cox et al. 1987, Kaczor and Hartnett 1990, Folkerts et al. 1993).
Gopher tortoise burrows are occupied by over 300 vertebrate and arthropod
species, some only living in tortoise burrows (Jackson and Milstrey 1988). Because of the importance of tortoise burrows to commensal animals, the gopher tortoise is considered a keystone species in sandhill ecosystems (Eisenberg 1983). Displacement of gopher tortoises by cogongrass invasion could mean the potential loss of these hundreds of commensal animals from sandhill ecosystems.
In a study in north-central Florida sandhill (Kalisz and Stone 1984), scarab beetles excavated an average of 3700 kg of soil in 6163 mounds/ha. Mounds were less numerous in sandhill with dense rather than sparse grass cover, thought by the authors to be due to obstruction of the beetles' flying and scavenging activities. In nearby sand pine scrub where subsurface and surface soils differed, scarab beetle mounds contained higher levels of P, Ca, Mg than the surrounding surface soil. In sandhill, scarab beetles mounds contained more clay particles, excavated from deep soils, than surrounding surface soils. In addition to altering surface soil chemistry and texture, the authors suggested that scarab beetle tunnels may provide root penetration pathways that would aid in acquisition of deeper soil water and nutrient resources.
Along with scarab beetles, pocket gophers influence soil genesis and soil profile morphology, preventing development of strong soil profiles because of their soil mixing activities, southeastern pocket gophers in a north-central Florida sandhill with a dense grass understory deposited an average of 8160 kg of soil in 558 mounds/ha (Kalisz and Stone 1984). Studies of other species of pocket gophers in the western United States
showed that mounds include feces and plant parts and therefore had higher organic matter than surrounding soils (Grinnell 1923). These pocket gopher mounds sometimes had higher levels of N, P, and K, and altered soil texture and water-holding capacity compared to surrounding soils. Western pocket gopher mounds typically contained more annual plant species than surrounding vegetation, and the vegetation changed as mounds aged (Huntley and Inouye 1988). southeastern pocket gopher mounds may similarly provide a range of resource conditions for seedling recruitment and growth in Florida sandhill.
Gopher tortoise mounds are not only larger than scarab beetle and pocket gopher mounds, but are also more compacted by years of regular soil excavation and tortoise passage across the mound to the burrow entrance (Hermann 1988). In a study in central Florida sandhill (Kaczor and Hartnett 1990), gopher tortoise mounds were significantly higher in pH and lower in organic matter and nutrients (except Na and P, which did not differ from surrounding soils). Tortoise mounds had higher light levels and greater diurnal temperature variation, and lower maximum temperatures during bumns. Abandoned mounds had higher densities of annual plant species than either recently-abandoned mounds or undisturbed soil. However, proportion of seedlings of a rhizomatous perennial, Pilyopsis graminifolia (silk grass), was higher on tortoise mounds than in undisturbed soil. At a density of 46 mounds/ha, gopher tortoise mounds were important sites for colonization of early successional species, maintaining an overall higher diversity of plant species than in undisturbed sandhill. In another study, germination of Bonamia grandiflora (Florida bonamia) in Florida scrub was highest after seed were buried in mounds (Hartnett and Richardson 1989).
Burrowing animals therefore likely influence sandhill ecosystem structure and function by soil mixing and mound formation. Decrease or loss of these species due to cogongrass invasion could result in changes in important soil mixing processes. Without mounds, opportunities for seedling recruitment, especially of annuals, may be reduced. To determine if cogongrass influenced the occurrence of sandhill burrowing animals, I measured frequency of mounds of these 3 species of sandhill burrowing animals in adjacent sandhill with and without cogongrass.
Since pocket gophers are important herbivores of below-ground plant tissue (Huntley and Inouye 1988), I graphically compared total non-structural carbohydrate content of roots and rhizomes of cogongrass and of sandhill plant species to preliminarily assess their nutritional value to fossorial herbivores.
Grazer/browsers. Florida sandhill are also habitat for numerous animals that graze on herbs and browse on leaves, twigs, and young shoots of trees and shrubs. These include Odocoileus virginianus (white-tailed deer), Sciurus niger shermani (Sherman's fox squirrel), Sylvilagusfloridanus (cottontail rabbit), Podomysfloridanus (Florida mouse), Sigmodon hispidus (cotton rat), and gopher tortoise (Myers and Ewel 1990). Wildlife diversity is highest in sandhill where high plant species diversity offers a seasonal assortment of food (Umber and Harris 1974). Cogongrass, in contrast, is known to be an inferior forage for livestock (Falvey et al. 1981). In order to assess the effect of cogongrass invasion on forage quality for grazing and browsing sandhill animals, I compared the nutritive value of cogongrass and of sandhill grasses, herbs, and twigs.
In summary, my purpose in this chapter is to examine invasion (spread) of cogongrass and consequent displacement of sandhill plants and burrowing and grazing/browsing animals.
I conducted these field studies in sandhill at the Citrus Tract of Withlacoochee
State Forest (see Figure 1-2) on the Brooksville Ridge in west-central peninsular Florida (Citrus County; lat 28065'N, long 8236'W). The forest ("Withlacoochee") consists of 16,600 ha of sandhill primarily managed for natural regeneration of longleaf pine and as slash pine plantations that are periodically clearcut and replanted. Forest managers conduct frequent prescribed bums for forest and game management, and are currently shifting from dormant-season to growing-season burning. I chose this location because it contains extensive infestations of cogongrass in relatively undisturbed sandhill vegetation with a recorded management history. Furthermore, the Florida Division of Forestry management staff were interested in my project and willing to implement bum treatments.
Additional vegetative spread data were obtained from a cogongrass infestation in sandhill on Riverside Island in Ocala National Forest (Marion County; lat 8249'N, long 2926'W), 97 km northeast of Withlacoochee. This site ("Ocala") was a private inholding within the National Forest and the property owners had requested that it not be burned with the surrounding sandhill forest. During the course of this study the National Forest bought the property and burned the site for the first time in 15 yr.
Voucher specimens of flowering cogongrass from these studies were submitted to the University of Florida herbarium (FLAS).
To estimate rate of vegetative spread of cogongrass, I measured horizontal extension of rhizomes and tillers in 7 treatments (Table 2-1).
Table 2-1. Seven treatments of cogongrass measured to assess lateral extension of rhizomes.
Cogongrass Location Fire Time Since Last Burn Sample
Type Management Size (n)
planted plugs sandhill fire-managed 5 yr 13
sward sandhill fire-managed 5 yr 13
sward sandhill fire-managed 5 yr then 1.5 yr 13
sward sandhill fire-excluded 15 yr 47
sward sandhill fire-excluded 15 yr then 0.3 yr 47
sward uncut plantation fire-managed 4 yr 12
sward clear-cut plantation fire-managed 4 yr 17
The planted cogongrass plugs each consisted of an intact 13-cm diameter core (15 cm deep) of soil, rhizomes, and tillers trimmed to 3 cm height. At 13 different sites across the forest, I dug a core from a random point within a dense cogongrass sward and transplanted it into sandhill vegetation burned 5 yr before. Transplants were watered (2 1 each) at planting. After 2 growing seasons I measured distance from the center of the plug to the farthest cogongrass tiller and calculated maximum rate of vegetative spread (m/yr) for each plug.
In the established cogongrass swards, metal stakes were placed at 10 m intervals at tillers on the edge of the sward in the dormant season. After 1 or 2 growing seasons, I measured the perpendicular distance from a 10 m line connecting adjacent stakes to the farthest cogongrass tiller extending beyond the edge of the sward. I then calculated maximum rate of vegetative spread (m/yr) for each 10 m interval.
1 constructed 4 models of cogongrass spread in fire-managed sandhill, projecting
the area covered after 20 years of growth. I assumed the slowest mean rate of spread
obtained from the studies described above for the first 2 yr of growth, increasing over the
next 4 yr to the mean of the 2 fastest rates of spread.
" Model I represents a sterile clone arriving in fire-managed sandhill as a seed or
rhizome, forming a single 0.25 m radius circular infestation and vegetatively spreading
outward evenly in all directions.
" Model 11 represents a sterile clone arriving as rhizomes spread along a forest road
edge in a 0.5 m x 10 m rectangle and vegetatively spreading outward into the forest
evenly from all sides except the road edge.
" Model III represents a low-fertility clone arriving and spreading as in Model 11. This
clone produces a small amount of viable seed that are wind-dispersed within 110 m of
the inflorescence, resulting in 1 new infestation (0.25 m radius circle) each year that
spreads as in Model 1.
" Model IV represents a high-fertility clone arriving and spreading as in Model 1. In
addition, this clone produces large amounts viable seed that are dispersed within 10
m of the inflorescence, resulting in 1 new infestation (0.25 m radius circle) for each
existing clone each year. Each new clone spreads as in Model I.
Displacement of Sandhill Vegetation
I conducted this study in 3 of the largest cogongrass, swards in sandhill in
Withlacoochee, in areas with similar management histories. Each sward extended at least
50 m along the road edge and at least 35 m into the adjacent sandhill. I divided sandhill
vegetation into 3 classes:
*herbaceous (grasses, forbs, and all seedlings)
* palms, shrubs, and small trees (woody plants < 1.2 m tall or < 2.5 cm diameter at 1.2
m height, "DBH")
" large trees (woody plants 2.5 cm DBH)
I estimated % cover of litter, cogongrass, and sandhill vegetation classes (except
trees 2.5 cm DBH) in 234 plots (1 m x I m) located at random points along the sward
edge and then systematically 6 m apart, extending 24 m into the cogongrass sward and 24
m into adjacent sandhill without cogongrass. I then measured the correlation between
cogongrass cover and cover of sandhill vegetation classes for all plots with at least 1% cogongrass cover. I also compared mean % cover of the sandhill vegetation classes in cogongrass with 50% or more cover and in sandhill.
For large trees (> 2.5 cm DBH), I measured DBH of all live trees in two 20 m x 260 m transects, one in cogongrass and the other in adjacent sandhill, no closer than 5 m to the edge of the cogongrass sward. I compared proportion and mean basal area of pine and of oak trees in cogongrass and in sandhill. Displacement of Sandhill Animals
Burrowers. I counted all mounds of scarab beetles, pocket gophers and gopher tortoises in March in two 2 m-wide transects in an area of Withlacoochee burned 2 mo before. Mounds were defined as patches of elevated bare soil with no emergent vegetation. Mounds < 10 cm diameter were assigned to scarab beetles, > 10 cm diameter to pocket gophers. Gopher tortoise mounds were identified by their burrow entrance. I performed this study in a recently-burned sandhill because mounds were readily visible. One plot followed the center of a roadside cogongrass sward that did not exceed 23 m width. The other plot roughly paralleled the first (30 m away) in adjacent sandhill without cogongrass. I skipped areas where the cogongrass was less than 10 m wide (or absent), pine needles concealed mounds, or the forest was converted to a quail food plot or drainage ditch. Total length covered for each treatment was 430 m.
To assess desirability of cogongrass rhizomes as food for pocket gophers, I compiled data from 2 studies and compared total non-structural carbohydrate (TNC) content in rhizomes of cogongrass (Gatfhey 1996) and in roots of wiregrass and 2 species of sandhill oaks (Woods et al. 1959). The cogongrass samples were collected monthly for
2 yr from a clay-settling pond in a phosphate mine in central Florida, and the sandhill samples were collected monthly for 3 yr from a sandhill forest in the western panhandle of Florida. I graphically compared monthly average TNC from both studies.
Grazer/browsers. To determine the desirability of cogongrass shoots as food for sandhill grazing and browsing animals, I analyzed the nutritional quality of above-ground biomass of cogongrass and of composite samples of sandhill grasses, herbs, and woody plants (upper 2.5 cm of shoots only). I collected samples from randomly located plots (1 m x 1 m) in an area burned in March (sampled 1 mo after burning) and in a nearby area burned in April (sampled at 3, 6, and 14 mo after burning). I removed dead leaves, leaf litter, pine needles and twigs, dried the samples to constant weight at 60 C, then ground them through a 1 mm sieve. The University of Florida Forage Evaluation Support Laboratory analyzed % total N, % total P, % total neutral detergent fiber, and ruminant in-vitro digestibility.
Throughout the remainder of this document, all statistical tests were analyzed at oc = 0.05. All-pairwise comparisons were considered significantly different at P < 0.05. Nonparametric tests were used if nonnormal data could not be transformed. Statistical power values reported for parametric tests are retrospective, i.e., calculated after the data was collected and analyzed (Thomas 1997).
Mean rate of maximum vegetative spread by tiller extension differed significantly (H = 50.1, df = 6, P < 0.01), ranging from 0.5 0.4 rn/yr for the planted cogongrass plugs to 2.6 0.9 m/yr for the cogongrass sward in fire-managed sandhill burned 1.5 yr before
(Figure 2-1). Southeast Asian studies also show that burning increased the tillering capacity of cogongrass (Goldammer and Penafiel 1990). For comparison, Mimosa pigra, a nonindigenous woody shrub with water-dispersed seed, spread a maximum of 76 m/yr in wetlands in northern Australia (Lonsdale 1993). Rate of spread for the cogongrass sward in fire-excluded sandhill (not burned in 15 yr) did not differ from the planted cogongrass plugs or the cogongrass sward in uncut pine plantation. Re-introduction of fire into the fire-excluded sandhill did not significantly increase spread rate after 0.3 yr. Spread rate in the clearcut pine plantation was significantly faster than in the uncut plantation, but did not differ from fire-managed sandhill. Recent burning in the fire-managed sandhill did not significantly increase spread rate.
Spread models (Figure 2-2) showed that after 20 yr, the sterile clones (Model I
and Model II) covered 0.5 ha (95% CI, 0.3 to 0.6 ha) and 0.5 ha (95% CI, 0.3 to 0.6 ha) respectively. It is notable that the circular infestation expanding on all sides exceeded the rectangular infestation (expanding on only 3 sides) in areal coverage after 14 yr. The lowfertility clone (Model III) invaded 3.3 ha (95% CI, 1.8 to 4.4 ha) and the high-fertility clone covered 4605.5 ha (95% CI, 1049.8 to 4646.1 ha) after 20 yr. The increase in rate of spread seen in Models III and IV is consistent with that demonstrated by a similar model of spread of an invading plant population with varying control of "satellite" populations (Moody and Mack 1988). Validated models of invasion of a tall annual herb, Impatiens glandulifera (Himalayan balsam), in the British Isles showed an areal coverage of 88,200 to 1,082,000 ha in 20 years (Perrins et al. 1993).
o 2- -AC44a
00 -. =-- 0,
Figure 2-1. Rate of cogongrass spread (m/yr) by extension of tillers for 7 treatments. Number in parentheses in x-axis labels is years since last burn. Box plots show mean (black square), median (center line), 75th and 25th percentile (top and bottom lines), and 90th and 10th percentile (top and bottom whiskers). Different letters indicate significant differences in pairwise comparisons (Dunn's method, P < 0.05),
,Model I (sterile circle) ...... Model II (sterile roadside rectangle) Model III (low-fertility, one new circle/yr)
-- Model IV (high-fertility, one new circle/circle/yr)
10000.0000 4605.5 ha
'-' 10.0000 ,3.3 ha
-" 1.0000 0. h
0 5 10 15 20
years since arrival
Figure 2-2. Four models of cogongrass spread in Florida sandhill, from arrival to 20 years, showing area (logarithmic scale) invaded over time. Rate of vegetative spread by tiller extension was estimated at 0.5 + 0.2 m/yr for the first 2 years and increasing over the next 4 years to 2.3 + 0.5 myr, based on the slowest and mean of the fastest rates of spread measured in sandhill.
Displacement of Sandhill Vegetation
Cover of all sandhill vegetation classes and litter were negatively correlated with cogongrass cover (Figure 2-3)--herbs most strongly (r. = -0.69, P < 0.01, n = 131), litter less so (r.= -0.48, P < 0.01, n = 131), and shrubs/palms/small trees even less (r, = -0.23, P < 0.01, n = 131). Median % cover (Table 2-2) was significantly lower in cogongrass (50% or more cover) than in sandhill for litter (T= 455.5, P < 0.01), sandhill herbaceous vegetation (T= 280.5, P < 0.01), and sandhill shrubs/palms/small trees (T= 816.0, P <
Table 2-2. Percent cover (median and upper and lower quartiles) of sandhill
vegetation and litter in 50% or more cogongrass cover (COG, n = 21) and in sandhill with no cogongrass (SAN, n = 109).
sandhill herbaceous sandhill shrubs/ litter
vegetation palms/small trees
COG SAN COG SAN COG SAN
median 1 36 0 5 9 31
lower quartile 0 23 0 0 6 17
upper quartile 4 56 1 23 14 54
I compared proportions of trees 2.5 cm DBH in 2 transects, one in cogongrass and one in sandhill (Table 2-3). There were significantly more longleaf pine in cogongrass than sandhill (z = 3.23, P < 0.01, power = 0.90). Proportion of oaks (Quercus incana, Q. laevis, Q. geminata) did not differ in cogongrass and sandhill (z = 0.99, P = 0.32, power =
0.17). Median basal area (Table 2-3) was higher in cogongrass for longleaf pine (T= 64512.5, P <0.01) and for oaks (T= 10261.5, P= 0.02).
-~ 60 r,=-0.69
0 ~ 0
0 10 20 30 40 50 60 70 80 90 100
S 60 r = -0.48
20 t + *. *
0 I I I III
0 10 20 30 40 50 60 70 80 90 100
6 r =-0.23
Q 40 -*
S 0 10 20 30 40 50 60 70 80 90 100
% cogongrass cover
Figure 2-3. Correlation between cogongrass cover (> 0%) and cover of
sandhill herbaceous vegetation (including seedlings), litter, and shrubs/palms/small trees (r. = Spearman rank order correlation coefficient).
Table 2-3. Sample size, proportion, and basal area (cm2) of longleaf pine and oak trees in cogongrass and sandhill.
longleaf pine oaks
cogongrass sandhill cogongrass sandhill
sample size 323 247 92 109
proportion 0.57 0.43 0.46 0.54
basal area (cm2):
median 188.5 131.7 39.7 20.4
lower quartile 85.1 49.5 20.4 10.1
upper quartile 339.6 277.5 77.1 106.1
Displacement of Sandhill Animals
Burrowers. In the cogongrass plot I counted 18 scarab beetle mounds, 82
pocket gopher mounds, and 1 gopher tortoise mound. In the sandhill plot I counted 80 scarab beetle mounds, 76 pocket gopher mounds, and 2 gopher tortoise mounds. Cogongrass had a smaller proportion of scarab beetle mounds than sandhill (z = 5.01, P < 0.01, power = 1.00). There was no difference in proportion of mounds of pocket gophers (z = 0.32, P = 0.75, power = 0.05) or of gopher tortoises (z = -0.69, P = 0.49, power =
Graphical comparison of mean total non-structural carbohydrates/mo for cogongrass rhizomes and for roots of wiregrass and 2 sandhill oaks showed that cogongrass had higher concentrations than any of the 3 sandhill species in all months (Figure 2-4).
Grazer/browsers. I compared mean % of total N, total P, fiber, and in-vitro
digestibility in cogongrass and sandhill forage at 1, 3, 6, and 14 mo after burning (Table 2-
- *-- wiregrass
--- turkey oak
0- bluejack oak 400
1 \ ---150 -..,
0 I I I I I 1F I
0 > .
Figure 2-4. Total non-structural carbohydrates (mg/g) of rhizomes from
cogongrass and of roots from 3 sandhill plant species. Cogongrass data from Gaffney (1996) sampled in a clay-settling pit in a phosphate mine in central Florida, 2 yr means and standard deviations. Sandhill data from Woods (1959) sampled from a sandhill forest in the western panhandle of Florida, 3 yr means and ranges of means for the 3 yr.
4, Figure 2-5). Mean % N and P were both significantly higher in cogongrass at 1 and 3 mo postburn, but were significantly lower in cogongrass at 6 and 14 mo postburn. Mean % fiber was lower in cogongrass at 1 and 3 mo postburn, was higher in cogongrass at 6 mo, and did not differ from sandhill at 14 mo. Mean % in-vitro digestibility was higher in cogongrass at 1 and 3 mo, was lower in cogongrass at 6 mo, and was higher in cogongrass at 14 mo. The sharp decrease in % fiber coupled with a sharp increase in invitro digestibility for sandhill samples collected 6 mo postburn, in October 1995, may be due to above-normal rainfall that month (67% or 9.88 cm above-normal) resulting in a flush of fall-flowering sandhill herbaceous species.
Table 4-2. ANOVA of 4 measures of nutritional quality of cogongrass and sandhill forage collected at 1, 3, 6, and 14 months after burning.
total df F P power
% N 61 4843.5 < 0.01 1.0
%P 61 811.8 <0.01 1.0
% fiber 61 139.4 < 0.01 1.0
in-vitro digestibility 61 400.3 < 0.01 1.0
Vegetative spread of cogongrass by extension of rhizomes and tillers was fastest for cogongrass swards in fire-managed sandhill and in clearcut pine plantation. Cogongrass spread rate was slowest for planted cogongrass plugs and for cogongrass swards in fire-excluded sandhill and in uncut pine plantation. Recent burning in both firemanaged and formerly fire-excluded sandhill tended to increase mean rates of spread, although not significantly. This was especially notable in sandhill where fire had been
2- 861.6- 821.2
\, t 780.8
0.4- 700 66 -g g I I I ,
1 3 5 7 9 11 13 1 3 5 7 9 11 13
135 7 91113
0.5- 450.4- .400.3 350
E 0.2- I 30
0.1 -------- -. 250 1 1 1 1 1I7 201 3 5 7 9 11 13 13 5 7 9 11 13
months since burn months since burn
Figure 2-5. Nutritional analysis of cogongrass shoots
("cogongrass") and of composite samples of sandhill grasses, herbs, and upper 2.5 cm of shoots of woody plants ("sandhill") collected 1, 3, 6, and 14 mo after burning Error bars represent 1 standard deviation from the mean. Differences between means for cogongrass and sandhill were statistically significant in a one-way ANOVA and Student-NewmanKeuls methods of all pairwise comparisons (P < 0.05) for all treatments except fiber at 14 mo postburnm, shown with an asterisk.
excluded for 15 yr; within 0.3 yr (4 mo) after reintroducing fire to this sandhill, maximum rate of spread leapt from 4.00 m/yr to 17.25 m/yr. Spread rate tended to be faster in firemanaged sandhill than in fire-excluded sandhill, evidence that management of sandhill forests by burning increases spread of cogongrass.
These data show that invasion by tiller extension alone (Models I and II) is relatively slow, 1 new infestation covering about 0.5 ha after 20 yr. If all cogongrass clones were sterile, then managers could prevent new infestations by assuring that road fill and equipment are free of cogongrass rhizomes. However, many swards probably produce at least small numbers of viable seed. This fits the low-fertility clone in Model III (producing 1 new clone/yr) which invades 7 times more area after 20 yr than the sterile clones. In Withlacoochee, roadside infestations rarely exceed 35 in in width in sandhill that has not been roller-chopped or clearcut in the past 20 yr. There are few infestations away from roadsides, although I did find a clone in interior sandhill radiating from an active gopher tortoise mound. This coverage suggests that the cogongrass population in this forest is intermediate between the sterile and low-fertility models.
Today's cogongrass in the southeastern United States may be the progeny of
relatively few founder genotypes. Clones of these founders may have remained separated by distances that did not allow for pollen mixing, resulting in few viable seed. Continued long-distance transport of cogongrass rhizomes across the Southeast could, therefore, result in mixing of clones that had been reproductively isolated. The high-fertility model (producing 1 new clone/existing clone/yr) indicates that cogongrass swards that produce viable seed every year can cover over 9000 times more area than sterile clones. If this becomes the rule instead of the exception, cogongrass will become uncontrollable.
Displacement of Sandhill Vegetation
As cogongrass cover increased, cover of sandhill vegetation (except large trees) significantly decreased. Similarly, in studies of African grass invasion in Venezuelan savanna, the total number of herbaceous plant species decreased after invasion from just over 100 to 2 species (Baruch 1996). Median % cover of sandhill herbaceous vegetation, including grasses and seedlings, was 97% lower in dense cogongrass than in sandhill without cogongrass. Cover of sandhill shrubs/palms/small trees was 100% lower and cover of litter was 71% lower in cogongrass than in sandhill (Table 2-2). The areas of dense cogongrass sampled in this study was no more than 24 m from the expanding edge of the cogongrass sward and are thought to be less than 20 yr old. This indicates that within 20 yr cogongrass displaced most of the existing herbaceous vegetation, shrubs, palms, and small trees.
The presence of more longleaf pines in cogongrass than in sandhill suggests that this species, once it reaches about 10.4 cm DBH (85.1 cm2 basal area, or lower quartile of surviving pines in cogongrass), is unaffected by cogongrass. These pines may continue to grow in spite of a solid understory of cogongrass, but the longleaf pine overstory is merely an artifact of pre-cogongrass conditions. Recruitment of seedling longleaf pine and other sandhill plants, naturally or by planting, is significantly reduced in cogongrass (discussed in Chapter 4). Both oak and pine trees in cogongrass had a larger mean basal area than oak and pine trees in sandhill. Since there is no evidence that cogongrass stimulates the growth of trees (resulting in larger trees in cogongrass), these data suggest that cogongrass contains fewer small diameter trees, possibly the result of mortality due to hotter
cogongrass fires (discussed in Chapter 3) or below-ground competition from cogongrass (discussed in Chapter 5).
Displacement of Sandhill Animals
Burrowers. There were 76% fewer scarab beetle mounds in cogongrass, perhaps due to obstruction of their flight by the relatively tall (to 1.5 m) cogongrass leaves. Significant reduction in numbers of this important agent of soil mixing could have long-term consequences for soil processes in sandhill, resulting in development of soil horizons typically absent in sandhill, that might in turn influence plant community composition.
Number of pocket gopher mounds did not differ between cogongrass and sandhill. The abundance of pocket gopher mounds in pastures, road sides, and lawns indicates that they are not deterred by dense rhizomes as found in cogongrass swards. Cogongrass may be a desirable food for pocket gophers because its rhizomes are high in sugars and starches, as are other closely-related grasses such as Saccharum officinarum (sugarcane) (Hunsigi 1986). In fact, cogongrass rhizomes were 30% to 65% higher in total non-structural carbohydrates than roots of wiregrass and 2 sandhill oaks (Figure 24). It has been shown, however, that pocket gophers feeding exclusively in Paspalum notatum (Bahia grass) pastures experienced a seasonal N deficiency that reduced reproduction (Ross 1976). Similar nutrient deficiencies might develop if pocket gophers were restricted solely to feeding on cogongrass rhizomes.
If indeed pocket gophers will continue to live in or on the edges of advancing cogongrass swards, as these data suggest, then they will continue to create soil mounds
that offer brief opportunities for sandhill seedlings to recruit and to escape from lethal burn temperatures.
Because of the low density of gopher tortoises in the study site, no conclusions about the effects of cogongrass on this species could be drawn. Habitat requirements for gopher tortoises, however, may not be met in dense cogongrass. Gopher tortoises require soils that allow easy burrowing, and cogongrass' dense rhizome layer (discussed in Chapter 5) may be a deterrent to burrowing for this species. Also, gopher tortoises require open areas for incubation of eggs, with female tortoises excavating nests in bare spots, usually in the apron outside their burrow (Cox et al. 1987). By rhizomatous growth, cogongrass would eliminate bare spots typical of sandhill's bunch grass vegetation, therefore reducing the number of sites suitable for tortoise nesting.
Additionally, the gopher tortoise diet consists of a seasonal diversity of herbaceous ground cover (Cox et al. 1987). In the spring and summer, they prefer high protein legumes and other dicots, and less than half their diet consists of grasses. In the fall and winter, when herbaceous dicots are less available, grasses make up 70% of their diet. When their diet is restricted to sandhill's low protein and high fiber grasses, gopher tortoises have lower growth rates. In dense cogongrass, there are almost no legumes or dicot herbs. Therefore, a gopher tortoise living in cogongrass would have to subsist primarily on cogongrass, which is generally even lower in N and higher in fiber than sandhill grasses, and would probably not provide the necessary nutrients for growth and reproduction.
I have found active tortoise burrows in dense cogongrass, with a narrow path of trampled cogongrass leading out to sandhill vegetation, usually < 10 m away. I could not
determine whether these were burrows maintained since cogongrass invaded the site or if they had been dug after cogongrass was present. I suspect that cogongrass' higher density and height compared to sandhill vegetation are deterrents to gopher tortoise movement. If cogongrass completely dominates a site, I expect that gopher tortoises would relocate or die. Since tortoise mounds are sites for seedling establishment and provide refuge from high fire temperatures for seedlings and young plants, the absence of tortoise mounds could affect plant community composition over time. Displacement of gopher tortoises by cogongrass invasion would also mean the potential loss of these hundreds of commensal animals from sandhill ecosystems.
Grazer/browsers. Cogongrass is nutritionally inferior as forage for sandhill's
grazing and browsing animals except in the first 3 mo after being burned. Cogongrass was significantly higher in N, P, and lower in fiber during the first month after fire. Three months after burning, nutritional quality of cogongrass declined but was still significantly higher than for sandhill vegetation. These results are consistent with findings in the highlands of Thailand (Falvey et al. 1981), where they harvested green leaves of cogongrass at 20 days (; 1 mo) and 120 days (; 4 mo) after cutting and found that % N declined from 2.93 to 0.56 and % P declined from 0.37 to 0.09. By 6 mo after burning, nutritional quality of cogongrass declined further and was lower than sandhill vegetation, and at 14 mo postburn, cogongrass was again lower than sandhill vegetation in N and P but did not differ in % fiber.
In addition to inferior nutritional value, cogongrass swards lack the seasonal
dietary variety of sandhill vegetation, since cogongrass displaces the diversity of grasses, herbs, and woody plants that provide a year-round assortment of food for sandhill animals.
Additionally, in areas north of the subtropics as in much of Florida, cogongrass itself is a seasonal food, unavailable after its leaves are killed by frost. Dietary needs aside, cogongrass' height is likely an impediment to many sandhill vertebrates. Cogongrass reaches 1.5 m in sandhill, 3 times taller than indigenous sandhill grasses (discussed in Chapter 3). This may make movement through cogongrass swards difficult for animals such as gopher tortoises and fox squirrels. I predict that sandhill animals that rely on the more open aspect of sandhill vegetation and feed on a diversity of plants species will gradually be displaced by cogongrass. This includes rare, threatened, and endangered animals, game species, and thousands of arthropod species that are important pollinators, herbivores, and detritivores in sandhill.
The purpose of these studies was to determine if cogongrass is a functional equivalent to sandhill grasses in ability to invade and thus displace other plants and animals. Comparisons of pertinent ecological traits showed that cogongrass spread vegetatively faster and farther than sandhill understory vegetation, potentially dominating thousands of hectares after 2 decades. As a consequence of invasion, cogongrass displaced most sandhill vegetation and some burrowing animals, and was inferior forage for grazing/browsing animals. These data support the hypothesis that cogongrass is not functionally equivalent to sandhill grasses. In subsequent chapters, I will examine the effect of cogongrass on the additional sandhill processes of fire regime, seedling recruitment, and resource availability.
COGONGRASS INVASION CHANGES SANDHILL FIRE REGIME Introduction
The purpose of the studies presented in this chapter is to ascertain whether
cogongrass is a functional equivalent to sandhill grasses in terms of the abiotic process of fire. I compared parameters of fire regime in sandhill with and without cogongrass. If these studies indicate that cogongrass significantly alters sandhill fire regime, this supports the hypothesis that cogongrass is not functionally equivalent to sandhill grasses. In this case, these data would provide evidence that cogongrass represents a new functional group in sandhill ecosystems and that this invader has the potential to exceed sandhill's resilience, shifting this ecosystem to a new stability domain, of grassland dominated by cogongrass.
Fire in Ecosystems
Recurrent fires are an integral part of the dynamics of many ecosystems (Pickett and White 1985, Trabaud 1987). Although ecosystems prone to frequent fire vary greatly, they are all characterized by periodic drought, frequent sources of ignition such as lightning, and nutrient-limited soils. Combustible fuels accumulate in these systems because vegetation growing on oligotrophic soils tends to be more flammable and produces litter that is more slowly decomposed than plants growing on fertile soils (Christensen 1987).
Fire maintains ecosystem structure by influencing production, succession, and species composition and diversity. Species richness is often highest immediately after burning. Fire also affects ecosystem function by changing nutrient cycles and water relations (Christensen 1981, Wright and Bailey 1982, Trabaud 1987). Variation in responses to fire results from differences in fire regime, determined by fire type and intensity, fire frequency, and season of burning (Sousa 1984, Pickett and White 1985, Malanson 1987). When a fire regime is changed, the ecosystem may consequently change in structure or function.
Invasive nonindigenous plants can change fire regimes or, in the case of invasive grasses, can initiate fire in ecosystems that were not formerly pyrogenic, resulting in the loss of indigenous species and alteration of important ecosystem processes, particularly nitrogen cycles (D'Antonio and Vitousek 1992, Hobbs and Huenneke 1992). Nonindigenous shrub invasions have altered fire behavior in South African fynbos shrublands (van Wilgen and Richardson 1985). In Australia, invasive grasses have altered fire regime and introducing fire cycles into non-pyrogenic tropical forest ecosystems (Humphries et al. 1991). Pennisetum polystachion (giant tussock grass) bums more intensely than indigenous Sorghum intrans in forests of wet tropical Australia (Macdonald and Frame 1988), while in arid regions of the continent Cenchrus ciliaris (bur grass) carries fires along watercourses that previously acted as fire barriers (Gill et al. 1990). Bromus tectorum (cheatgrass), one of the most extensive invasive plants in North America, is a highly flammable European annual that displaced less-flammable perennial grasses across the intermountain region of western North America in conjunction with livestock grazing (Mack 1986). Cheatgrass increased the size and frequency of fires,
leading to increased flooding and erosion, and displacement of shrubs and indigenous grasses (Klenunendson and Smith 1964). In Hawaii, invasion of grasses such as Andropogon virginicus (broomsedge), Me/mnis minutiflora (molasses grass), Schizachyrium condensatum (bush beardgrass), and Pennisetum setaceum (fountain grass) have increased the frequency and size of fires in woodlands (Hughes et al. 1991, Smith and Tunison 1992). In submontane Hawaii, molasses grass has converted nonflammable indigenous woodlands into highly flammable grassland (Hughes et al. 1991).
As mentioned earlier, fire regime is determined by several parameters: fire type and intensity, fire frequency, and season of burning (Sousa 1984, Pickett and White 1985, Malanson 1987). Fire type, the first parameter of fire regime, describes either surface or crown fires, and backing fires (spreading into the prevailing wind) or head fires (spreading with the prevailing wind). Head fires move rapidly, burn less fuel, and release less heat over a shorter time than backing fires. Fire intensity is an expression of energy released per unit length of burning front per unit time, and is a function of abiotic factors such as weather and topography as well as fuel characteristics such as quantity, structure, and flammability (likelihood of ignition). Low tissue nutrient concentration slows decomposition and thus increases fuel quantity. Flammability, largely determined by the chemical composition of plant tissues, is increased by low tissue water content.
Fire temperature is a surrogate indicator of fire intensity and a predictor of plant tissue death. In grasslands and savannas, temperatures at the soil surface range from 83 TC to over 1000T0 during fires (Wright and Bailey 1982). Below the soil surface, temperatures during a fire decrease sharply with depth, rarely exceeding about 80T0 at 0.6 cm depth except under burning wood. Plant tissue death during a fire occurs
instantaneously at about 66 C for unprotected tissues. Since combustion occurs when fuel temperature reaches 346 C 40 'C, protected meristematic tissues can be exposed to lethal temperatures when attached leaves, pine needles and bark burn (Albini 1993). Fire temperatures above 346"C are, therefore, an predictor of damage or death even of protected meristematic tissues.
Fire frequency, the next parameter of fire regime, is the return time at a particular location or the number of fires per unit area over a landscape. Fire frequency is determined by availability and regularity of ignition, weather favorable for ignition, net primary productivity, and rate of decomposition. Season of burning, the last parameter of fire regime, is determined by availability of ignition sources and seasonal variation in fuel moisture and quantity.
Sandhill Fire Regime
Fire has long been recognized as a important process in maintaining sandhill's
characteristic structure and function (Christensen 1981, Myers 1990, Robbins and Myers 1990, Streng et al. 1993). The natural fire regime of sandhill consists of light to moderate intensity surface fires, probably occurring every 2 to 8 years, and fueled primarily by perennial caespitose grasses such as wiregrass and pineywoods dropseed, but also by pine needles and oak leaves. Lightning-ignited fires occur during the spring and summer when thunderstorms are most common. There is also evidence that indigenous Americans also burned sandhill forests.
When burning is prevented in sandhill, the herbaceous understory and pine
overstory is displaced by xeric or mesic hardwood forest, as has happened with much of the southeastern sandhill ecosystem (Christensen 1981). Since European settlement many
pine forests in the Southeast have been burned in the dormant season for silvicultural reasons and for game management (Myers 1990, Robbins and Myers 1990). Present-day land managers are switching to growing season fires for ecological reasons, mostly because it is now known that growing season fires trigger flowering and seed production of many herbaceous species in the understory, notably grasses (Streng et al. 1993, Anderson and Menges 1997).
Longleaf pine, the dominant species in the sandhill overstory, is the most fireresistant pine species in the Southeast (Grace and Platt 1995). Fire can either enhance or eliminate this species, depending on fire regime. Since it requires sparsely vegetated or bare mineral soil to successfully establish, longleaf pine seedling recruitment is highest when the large seed with its persistent wing is able to reach bare ground soon after burning of sandhill (Croker and Boyer 1975, Landers 1991). Recently burned sites also provide a few years of protection from fire for pine seedlings, until fuels accumulate. After germination longleaf pines exist for 3 to 5 years in a "grass stage" during which the belowground apical meristemn is protected from low intensity fires. Then the juvenile "bolts" over
3 to 4 years, rapidly gaining in height and exceeding the understory grass layer where lethal fire temperatures are more likely to occur (Maple 1975). Thickening bark on developing pines increasingly insulates and protects the trunk from damage during low intensity fires (Wade 1986, Boyer 1993). Fire prevention and subsequent fuel accumulation prevent longleaf pine recruitment and increase the possibility of fire temperatures lethal even to large longleaf pines (Christensen 198 1, Wade and Johansen 1986, Myers 1990).
I hypothesized that cogongrass invasion alters sandhill fire regime primarily by changing characteristics of fine fuels (grass, litter and duff less than 0.6 cm in diameter), the most influential factor in the ignition and spread of sandhill fire (Wade 1989). 1 predicted that cogongrass fine fuels would be greater in mass, height, horizontal continuity, and flammability than sandhill fine fuels, and would accumulate more rapidly after burning. Consequently, cogongrass fires would have higher maximum temperatures at greater heights, would spread faster, and burn more evenly and more frequently than sandhill fires. I also predicted that, because of higher fire temperatures at greater heights, burned longleaf pine juveniles in cogongrass would have decreased growth and survival compared to those in sandhill.
This study was conducted in the Citrus Tract of the Withlacoochee State Forest ("Withlacoochee") (Citrus County; lat 2865'N, long 8236'W) in sandhill dominated by cogongrass ("cogongrass") and in adjacent sandhill with indigenous sandhill plant species ("sandhill") that had not burned in 4 years. Four prescribed fires were conducted by Withlacoochee State Forest staff in the early growing season: the first 3 in April 1995 and the 4th in March 1996. Each fire was ignited as a backing fire lit across the cogongrass and adjacent sandhill, with wind speed less than 16 km/hr. Shifts in wind direction after ignition caused the third and fourth fires to become head fires instead of backing fires.
I compared several prefire fine fuel characteristics in cogongrass and sandhill. I estimated fine fuel horizontal continuity by measuring the total length of bare ground under randomly located 100-m line transects (n = 13). I estimated other fine fuel attributes in randomly located 1 m x 1 m plots. For fine fuel mass, I harvested fine fuels in each plot
(n = 13), then oven-dried them at 60'0C to constant weight before weighing. To determine fine fuel vertical distribution, I harvested fine fuels from 3 height classes (0 to 0.49 m, 0.50 to 0. 99 m, and 1. 00 to 1. 50 m) (n = 13), then oven-dried them at 60 TC to constant weight before weighing. For fine fuel water content, I harvested fine fuels (n = 10), stored them in sealed plastic bags, weighed them the next day (fresh weight), then oven-dried them at 60 TC to constant weight before weighing again (dry weight). To measure heat of combustion (kJ/g), I oven-dried leaves of cogongrass and of sandhill grasses (n = 9) at 60 0C to constant weight, then ground and sifted them through a 1 mm sieve before a subsample was oxidized in an oxygen bomb calorimeter (Parr Model 13 41, Parr Instrument Company, Inc., Moline, Illinois, USA).
During each of the 4 fires, I measured maximum temperatures at 0, 0.5, and 1.5 m height at randomly located points in cogongrass and in sandhill. Eighteen temperatureindicating paints (Tempilaq@, Tempil Division, Air Liquide America Corporation, South Plainfield, New Jersey, USA), each formulated to melt at a specific temperature, were applied in a row to narrow rectangular steel strips (1.9 cm x 30.5 cm) attached horizontally to a metal pole, 1 strip at each of the 3 heights (n = 13). Temperatures tested
('C) were 94, 122, 150, 178, 206, 234, 262, 290, 318, 346, 374, 402, 430, 458, 542, 654, 766, and 878. These spanned the range of published values for grassland and savanna fire temperatures (Daubenmire 1968, Frost and Robertson 1985). After the fires, melted paint spots were recorded for each strip. Rate of spread of the fire front was recorded between two 1.8 in-tall metal poles placed at the edges of the burned areas, ranging from 35 to 145 m apart.
I calculated fire intensity using the following formula (Byram 1973): I = HWR
where I = fire intensity in kW/m; H = heat of combustion in kJ/g; W = fuel mass in g/m2; and R = rate of fire spread in m/sec. I used mean heat of combustion, mean fuel mass, and rate of spread for each fire.
To compare postfire fine fuel accumulation rates in cogongrass and in sandhill, I randomly selected 1 m x I m plots in cogongrass and sandhill (n = 13) at 3, 6, and 14 mo after burning. I harvested fine fuels in each plot and oven-dried them at 60 C to constant weight before weighing.
To compare effects of fire on survival and growth of established longleaf pine
juveniles (grass stage to 1.5 m tall), I tagged and measured the height (to top of meristem) and basal stem diameter of existing longleaf pine juveniles of various heights in a naturally regenerating stand, 150 in cogongrass and 150 in sandhill, in an area burned less than 1 month before. One year later I relocated the tagged pines and remeasured height and stem diameter of live pines.
Sandhill fine fuels were less horizontally continuous, with 3.0 % total bare ground compared to 0.3 % in cogongrass (T= 134.5, P = 0.04). Cogongrass had a significantly higher prefire fine fuel mass than sandhill (Figure 3-1, t = 3.48, df= 24, P < 0.01, power = 0.91). Fuel mass for sandhill in this study was consistent with a study that showed a steady state fine fuel mass of 750 g/m2 in sandhill 4 to 5 years after burning (Parrott 1967). Fine fuel mass for cogongrass in this study was higher than measured for a variety of indigenous neotropical savanna grasses, although it was similar to fine fuel mass of
cogongrass 1400 ----- sandhill
16 1000 o 800
0I I I I I
O\ 0 ,- t- 0 e D \O O "l N r W W
months since fire
Figure 3-1. Mean fine fuel mass accumulated at 3, 6, 14, and 48 mo after
burning in cogongrass and in sandhill. Error bars represent 1 standard deviation from the mean. Asterisks denote significant differences between cogongrass and sandhill at alpha = 0.05 level.
Hyparrhenia rufa, an African grass that invades and dominates these savannas (Sarmiento 1984). Mass of fine fuels was not different between cogongrass and sandhill from 0 to
0.49 m height (Figure 3-2, T= 211.0, P = 0.07), but was higher in cogongrass from 0.50 to 0.99 m (T= 260.0, P < 0.01) and from 1.00 to 1.50 m (T= 260.0, P < 0.01). Mean % water content in fine fuels did not differ between cogongrass (46.2 10.7) and sandhill (42.3 12.3) (1 = 0.75, df= 18, P = 0.46, power = 0.05). Mean heat of combustion was slightly higher for sandhill grasses (18.4 kJ/g 0.20) than for cogongrass (18.7 kJ/g
0.22) (t = -3.37, df =16, P < 0.01, power = 0.87). These were within the range of published values of heat of combustion for similar grassland fuels (Albini 1993, Glitzenstein et al. 1995).
There was no difference in rate of spread during the fires (Table 3-1, W= -2.00, P = 0.75). Data did suggest a trend toward faster rate of spread of head fires in cogongrass. Fire intensity also did not differ between cogongrass and sandhill (paired t =-1.55, df = 3, P = 0.22, power = 0.14). There was, however, a trend toward higher fire intensity values for cogongrass, especially for head fires. Fire intensity values were consistent with published values for similar grassland ecosystems worldwide (Frost and Robertson 1985).
Table 3-1. Rate of spread and fire intensity for each fire in cogongrass (COG) and in sandhill (SAN). Fire intensity was calculated using the formula I = HWR, where H equals mean heat of combustion for cogongrass (18.4 kJ/g) and sandhil (18.7 kJ/g), W equals mean prefire fine fuel mass for cogongrass (1163.2 g/m2) and sandhill (776.9 g/m2), and R equals rate of spread for each of the 4 fires.
R = rate of spread (m/sec) I = fire intensity (kW/m)
Fire # Fire Type COG SAN COG SAN
I backing 0.0185 0.0235 395.95 341.41
2 backing 0.0208 0.0195 445.18 283.30
3 head 0.1300 0.1300 2782.37 1888.64
4 head 0.1458 0.0280 3120.54 406.78
Ms cogongrass Ssandhill 800
0 -.49 .50-.99 1.00-1.50
height class (m) Figure 3-2. Mean fine fuel mass in 3 height classes in cogongrass and in sandhill. Error bars represent 1 standard deviation from the mean. Asterisks denote significant differences between cogongrass and sandhill at each height (P < 0.05).
For all fires combined, proportion of strips at 0.5 and 1.5 m that did not reach 94 C (the temperature at which the first paint spot would melt) was higher in sandhill than in cogongrass (Table 3-2). Cogongrass tended to have the highest temperatures at each height in each fire (Table 3-3).
Table 3-2. Proportion of strips at each height (all fires combined) with temperature < 94 C (n = 52).
Height Cogongrass Sandhill Test Statistic P power
0 m 0.00 0.06 Fisher Exact 0.24 -0.5 m 0.02 0.26 X2 = 10.1 < 0.01 0.91
1.5 m 0.12 0.53 X2 = 16.8 < 0.01 0.99
Table 3-3. Highest temperature (C) recorded for each height for each of the fires (n = 13) in cogongrass and in sandhill. During fire #4, no strips reached 94 0C at 1.5 m height in sandhill.
0m 0.5m 1.5m
Fire # Fire Type COG SAN COG SAN COG SAN
1 backing 318 290 290 290 262 150
2 backing 290 290 290 262 290 262
3 head 318 290 262 290 458 290
4 head 458 318 458 150 346 < 94
There was a significant difference in mean maximum temperature (Table 3-4) between cogongrass (260.9 13.7 0C) and sandhill (218.3 14.5 'C), as well as a significant effect due to height. There was no interaction between vegetation and height. At individual heights, however, there was no detectable difference between cogongrass and sandhill (Figure 3-3). Mean maximum fire temperatures in this study were similar to those measured in a study in central Florida sandhill (Williamson and Black 1981) and were within the range for other grassland fuels (Wright and Bailey 1982).
350 mom E0.5 m N 1.5 m
300 a a
Figure 3-3. Mean maximum fire temperatures at 3 heights in cogongrass and in sandhill for all 4 fires combined. Error bars represent 1 standard deviation from the mean. Different letters represent significant differences at P < 0.05.
Table 3-4. ANOVA of mean maximum temperatures at 3 heights in cogongrass and sandhill.
Source of Variation df F P power
vegetation 2 4.57 0.05 0.42
height 1 4.66 0.02 0.59
vegetation x height 2 0.98 0.40 0.05
After burning, significantly more fine fuel had accumulated in cogongrass than in sandhill at 3 mo (Figure 3-1, T= 242.0, P < 0.01), 6 mo (t = 4.97, df= 24, P < 0.01. power = 1.00), and 14 mo (t = 3.34, df= 24, P < 0.01, power = 0.88).
Percent mortality of longleaf pine juveniles burned 1 yr before (Figure 3-4) in cogongrass and in sandhill was not significantly different for pines 0 to 0.49 m tall (z =
1.50, P = 0.13, power = 0.32) or for pines 1.00 to 1.50 m tall (z = -0.48, P = 0.68, power = 0.08). Mortality was significantly higher in cogongrass for pines from 0.50 to 0.99 m tall (z = 1.98, P = 0.05, power = 0.51). Height growth in the 1 yr since burning for pines 0 to
0.49 m tall was significantly lower in cogongrass (21%) than in sandhill (50%) (T= 6162.5, P < 0.01). There was no difference between cogongrass and sandhill in height growth for pines 0.50 to 1.50 m tall (t = 0.18, df= 19, P = 0.86, power = 0.05). Stem diameter growth was not different between cogongrass and sandhill for pines 0 to 0.49 m tall (T= 666.5, P = 0.42) or for pines 0.50 to 1.50 m tall (T= 27.5, P = 0.15).
Cogongrass significantly changed fine fuel characteristics that determine sandhill fire regime. There was 90% less bare ground in cogongrass, so fine fuels were more evenly distributed horizontally. Consequently, cogongrass fires were more continuous and
80 sandhill 20 6
70 60 50
0 -.49 .50-.99 1.00- 1.50
juvenile pine height class (m)
Figure 3-4. Percent mortality of burned juvenile longleaf pines in 3 height
classes 1 yr after burning in cogongrass and in sandhill. Number of pines in each height class shown at the top of each column. Asterisk denotes significant difference between cogongrass and sandhill at alpha = 0.05.
all temperature strips at ground level in cogongrass reached at least 94 C. In contrast, sandhill fires were more patchy and 6% of the strips at ground level in sandhill did not even reach 94 C. Other invasive grasses have been shown to area extent of fires because of greater continuity of fuels (Klemmendson and Smith 1964, Gill et al. 1990).
Cogongrass had 50% more fine fuel by weight than sandhill at the time of burning. Consequently, cogongrass was able to reach higher maximum fire temperatures than sandhill at all heights. Cogongrass fuel was distributed higher above the ground: 27% of the fine fuels were above 0.5 m, compared to only 8% in sandhill. Cogongrass fuels from
0.50 to 0.99 m height were 4.5 times greater in mass than sandhill fuels at this height. From 1.00 to 1.50 m height, cogongrass fuels were 6 times greater in mass than in sandhill. As a result, cogongrass averaged 50 C hotter at 0.5 m and 73 C hotter at 1.5 m, although these are not statistically significant differences due to the high variance in fire temperatures. Greater shoot biomass of nonindigenous grasses contributed to hotter fires in other Australian ecosystems (Macdonald and Frame 1988).
Fire temperatures in cogongrass reached a high of 458 C on some strips at all heights. Since bark combusts at 346 C and unprotected plant tissue is instantaneously killed at 66 0C, temperatures of 4580 C could result in significant loss of above-ground tissues of fire-tolerant sandhill plant species. If these lethal temperatures persist for more than the few seconds of a passing flame front, even thick-barked trees could be killed. Data on fire temperature duration would help to determine direct mortality effects of cogongrass fires on sandhill vegetation (Jacoby et al. 1992).
After burning, cogongrass fine fuels recovered more quickly than sandhill fuels. In
Southeast Asia, cogongrass is also known to recover quickly from fire (Eussen 1980). In this study, cogongrass had over 100% more fine fuel than sandhill at 3 mo after burning, 86% more at 6 mo, and 50% more at 14 mo. With twice as much fuel after only a few months of growth, fire in cogongrass could ignite and spread more frequently in the absence of fire management. Greater fire frequency due to nonindigenous grass invasion has been documented in ecosystems worldwide (Klemmendson and Smith 1964, Hughes et al. 1991, Humphries et al. 1991, Smith and Tunison 1992).
In both sandhill and cogongrass, the bulk of fine fuel is below 0.5 m; consequently, pines smaller than 0.5 m showed similar levels of mortality after fire in both fuels. From
0.50 to 0.99 m, however, cogongrass contained 4.5 times more fuel than sandhill (274 + 63 g/m2 compared to 59 27 g/m2) and the mean maximum temperature was 50 'C hotter than sandhill at 0.5 m height. Correspondingly, mortality of burned pine juveniles 0.50 to
0.99 m tall was 31% higher in cogongrass. In sandhill, rapid growth out of the grass stage offers longleaf pine juveniles some protection from fire. In contrast, longleaf pines juveniles in cogongrass are still vulnerable to fire damage from 0.50 to 0.99 m height because of the greater quantity of fuel and higher fire temperatures. In spite of cogongrass' greater fuel mass from 1.0 to 1.5 m height, and greater mean maximum temperatures at
1.5 m, fuel mass was still relatively low (23 10 g/m2) from 1.0 to 1.5 m height, and mortality of pine juveniles did not differ between cogongrass and sandhill.
A similar study in northern Thailand (Koskela et al. 1995) examined Pinus
merkusii, which grows on nutrient-poor, well-drained soils in association with cogongrass, much like longleaf pine in southeastern sandhill. This Asian pine exists for several years in a grass stage that is resistant to drought and fire, again like longleaf pine. Although Pinus
merkusii germinated well under dense cover of cogongrass, seedlings were then killed in frequent fires fueled by cogongrass, so that survival of pines beyond the grass stage was found to be highest when cogongrass was burned just before pine seed fall.
Burned longleaf pine juveniles 0 to 0.49 m tall showed lower height growth in
cogongrass compared to sandhill. In Chapter 4, I will discuss resource-limitation studies which suggest that reduction in seedling growth may more likely be due to competition from cogongrass instead of direct fire effects.
Cogongrass changed fire regime in sandhill by significantly changing fine fuel mass, horizontal and vertical distribution, and rate of accumulation. These changes resulted in hotter, higher fires that burn more continuously and more frequently. Over time, this could result in higher mortality of seedling and juvenile sandhill plants and of less fire-tolerant sandhill species. In terms of fire, cogongrass is not an ecological equivalent of sandhill grasses.
COGONGRASS INVASION REDUCES SANDHILL SEEDLING RECRUITMENT Introduction
The purpose of this chapter is to determine whether cogongrass is a functional equivalent to sandhill grasses in terms of seedling recruitment of indigenous sandhill species. I compared survival and growth of planted seed and seedlings of several sandhill species representing various life-forms. If these studies indicate that cogongrass significantly reduces seedling recruitment in sandhill, this supports the hypothesis that cogongrass is not functionally equivalent to sandhill grasses. In this case, these data would provide additional evidence that cogongrass represents a new functional group in sandhill ecosystems.
Seedling recruitment by sexual reproduction is often the crucial life-history stage for population persistence (Harper 1977, Goldberg 1982, Gross and Werner 1982, Peart 1989) and is critical to maintenance of species diversity (Grubb 1977, Tilman and Pacala 1993). Adult-seedling interactions are a key factor in analysis of the population basis of community structure and succession (Peart and Foin 1985). Sexual reproduction in plants involves production of viable seed, dispersal in space and time, seedling recruitment (the focus of this chapter), and growth beyond the seedling stage (Harper 1977). Seedling recruitment is strongly coupled to disturbance for many plant species with seeds that
require gaps in vegetation cover to germinate and establish (Silvertown and Doust 1993). For these species, disturbances that reduce vegetation cover, such as fire, tree fall, lightning strike, disease outbreak, and ungulate grazing, provide opportunities for seedling recruitment, which subsequently influences the size, structure, and species diversity of plant communities (Sousa 1984). In studies of various herbaceous and woody species for which sexual reproduction is limited to a I to 5 year period following disturbance, recruitment was highest for species with larger seed that arrived soon after disturbance (Canham and Marks 1985).
Seedling recruitment is also influenced by life-form, i.e., single-character-based functional groups (Solbrig 1993. Life-forms might be grouped by such traits as longevity (annual vs. perennial, herbaceous vs. woody), architecture (tree, shrub, grass, etc.), lateral spread (clonal vs. non-clonal, rhizomatous vs. caespitose), association with symbiotic microbes, or seed size. For example, in a study of seedling recruitment of various plant life-forms in a N-limited prairie, large-seeded N-fixing legumes established most successfully [Tilman, 1997 #121). Large-seeded plants have seed reserves that allow seedlings to survive in harsher physical environments than small-seeded species (Tilman 1988, Westoby et al. 1996). Consequently, large-seeded species generally have a higher recruitment rate than small-seeded species under deeply shaded, xeric, or oligotrophic conditions (Grime and Jeffrey 1964, Canham and Marks 1985, Silvertown and Doust 1993).
Nonindigenous grasses have been shown to impede seedling recruitment of indigenous species, both woody and herbaceous. In California's valley grasslands, introduced annual grasses reduced recruitment of indigenous Quercus douglasii (blue oak)
seedlings (Gordon et al. 1989). In Texas, nonindigenous Cynodon dactylon (Bermuda grass) reduced growth of seedlings of three indigenous woody species, Baccharis neglecta (eastern baccharis) (Van Auken and Bush 1990b), Acacia smallii (huisache) (Cohn et al. 1989), and Prosopis glandulosa (honey mesquite) (Bush and Van Auken 1987, Van Auken and Bush 1990a). In rangeland in the Great Plains region of the United States, where nonindigenous annual grasses such as Bromus tectorum (cheatgrass) now dominate, revegetation studies using indigenous perennial grasses showed that intense competition from the invasive annuals, especially for soil moisture, prevented seedling establishment (Evans and Young 1972).
Seedling Recruitment in Sandhill
As in many other ecosystems, seedling recruitment maintains sandhill's notably high plant species diversity and is essential for long-term persistence of pyrogenic grass and pine species. Indigenous sandhill plant species cover a wide range of life-forms and reproductive strategies, in an ecosystem that is xeric, oligotrophic, and prone to frequent fires. Several of the common woody species in Florida sandhill, including Quercus geminata (sand live oak), Quercus incana (bluejack oak), Myrica cerifera (wax myrtle), and Diospyros virginiana (persimmon), reproduce vegetatively by clonal growth as well as by seed. Clonal growth is defined here as the horizontal extension of a plant by the addition of ramets that develop their own roots (Silvertown 1987). Sandhill is dominated, however, by species that reproduce primarily by seed, such as Aristida stricta (wiregrass), Sporobolusjunceus (pineywoods dropseed), and Pinuspalustris (longleaf pine). Since these are "driver" plant species in the functional group that perpetuates fire in sandhill
(discussed in Chapter 1), seedling recruitment in sandhill is essential to the long-term dominance of pyrogenic species.
In sandhill, growing-season fires stimulate flowering and seed production for many dominant sandhill grasses, notably wiregrass, and forbs (Lemon 1949, Parrott 1967, Clewell 1989, Platt et al. 1991, Streng et al. 1993). In addition, immediately following fire, conditions are favorable for successful establishment of many sandhill species (Carrington 1996). Burning provides opportunities for seedling recruitment by eliminating the litter layer and exposing bare ground, releasing nutrients to the soil, and reducing above-ground competition. Studies of recently burned sandhill found that soil levels of some plantavailable nutrients (P04, K+, Ca++, and Mg +) were higher and above-ground competition was reduced (Christensen 1977). In sandhill burned just before seed fall, seedling recruitment increased for longleaf pine and several herbaceous species (Lemon 1949, Platt et al. 1988). Seedling recruitment of many herbaceous species, including grasses, was also higher in sandhill burned in the growing-season (Whelan 1985, Brewer and Platt 1994, Anderson and Menges 1997). In summary, burning sandhill in the growing-season enhances seed production and recruitment of sandhill grasses and longleaf pines, the firecarrying species in sandhill, maintaining pyrogenic species and species diversity.
As discussed in Chapter 2, sandhill ecosystems in the southeastern United States are being invaded by cogongrass, which has been clearly shown to reduce seedling recruitment and arrest succession in deforested areas in Southeast Asia, India, Thailand, and the Philippines (Hubbard et al. 1944, Brook 1989, Turvey 1994). When cogongrass invades and dominates sandhill, it displaces extant understory vegetation and forms a taller, more massive understory (discussed in Chapters 2 and 3). In order for sandhill plant
species to persist with cogongrass, they must be able to recolonize via seedling recruitment in the cogongrass understory. Although burning before seed fall enhances seedling recruitment in sandhills, this may not be the case in cogongrass, which resprouts more rapidly after fire than indigenous sandhill herbs (see Chapter 3). In this chapter, therefore, I examine the effect of cogongrass and of burning before planting on recruitment of planted seed and seedlings in Florida sandhill.
These studies were conducted in the Citrus Tract of Withlacoochee State Forest ("Withlacoochee") in sandhill dominated by cogongrass ("cogongrass") and in adjacent sandhill dominated by indigenous sandhill plant species ("sandhill"). I compared survival and growth for planted seed and/or seedlings of 7 indigenous sandhill plant species ("sandhill seedlings") representing various plant life-forms (Table 4-1). These were planted in 4 treatment combinations: cogongrass or sandhill, not burned in 4 years ("unburned") or burned in April, 2 months before the beginning of this study ("recently burned"). Since seed and greenhouse-grown seedlings were planted, these studies measure recruitment success when barriers to seed dispersal and germination were overcome.
Seed of 6 species (Dalea pinnata, Pityopsis graminifolia, Sorghastrum secundum, Myrica cerifera, Serenoa repens, and Quercus geminata) were collected from at least 12 individuals at or near Withlacoochee during the fall. Seed were cleaned if necessary, and a portion of the seed of each species was immediately planted in the greenhouse in tubes (5.4 x 14.0 cm) filled with potting soil. The remainder was air-dried overnight and refrigerated at 7 oC until field planting in June.
0 0 0
0 od 40. 10. to 0.
0 0 0
Cd Cd _61 451
0 0 0 0
0 0 0
0 ed Cd Cd Cd u u Cd u
Cd Cd r.
Cd Cd r.
z ts ;:l Cd ;m 0
u _Q IT, =0
Thirteen replicates of each of the 4 treatments (cogongrass or sandhill, unburned or recently burned) were installed in June. Each replicate consisted of a randomly located 6 m-long line. At points spaced 0.1 m apart along the line, I planted 2 seed of one of the larger-seeded species (Quercus and Serenoa) or 10 seed of one of the smaller-seeded species (Dalea, Pityopsis, Sorghastrum, and Myrica). There was a total of 10 planting points for each of the 6 species, each point marked with a wire inserted in the ground. Seed were inserted just below the soil surface using long forceps, with minimum disturbance of the litter layer. Seed were not watered at planting time. If planting occurred on a mound created by a burrowing animal, then size and condition of the mound was noted. After 2 growing-seasons, I harvested all above- and below-ground biomass ("shoots" and "roots") of surviving seedlings, which were oven-dried to constant weight at 60 C to determine dry weight. Survival and growth of surviving seedlings was then compared.
Thirteen replicates of each of the 4 treatments were installed in June. Each
replicate consisted of a randomly located plot (2 x 10 m) divided into 20 subplots (1 x 1 m). In each replicate I planted 4 seedlings of each of 5 species (Pityopsis, Sorghastrum, Myrica, Quercus, and Dalea), one seedling per subplot. Seedlings of each species were identical in age, container size, height and stem number. Dalea was planted in only 5 replicates in 2 treatments (unburned cogongrass and unburned sandhill) because of small numbers of available seedlings. The base of each Sorghastrum seedling was loosely tied with flagging tape to distinguish it from surrounding grasses. Seedlings were watered (2 1
each) at planting. After 2 growing-seasons, I harvested all above-ground biomass ("shoots") of surviving seedlings, which were oven-dried to constant weight at 60 TC to determine dry weight. Root mass was not compared because most of the seedling roots were indistinguishable from cogongrass roots.
In order to compare recruitment success among various plant life-forms
(regardless of fire effects), I combined burned and unburned treatment data and compared recruitment success among the 5 species of planted seedlings. I standardized shoot growth data using the formula:
Z = (value sample mean) x sample standard deviation'
By rule, the standardized data have a mean of 0 and a standard deviation of 1. I added 1 to each standardized data point for ease of graphing.
Pinus seedling recruitment was examined in a separate study in which I measured survival and growth of planted pines in a 4 year-old sandhill pine plantation in Withlacoochee. Forest records (Andrea Crisman, personal communication) showed that before it was clear-cut in 1987, the 16 ha plantation consisted of dense rows of pines with a narrow band of cogongrass established along the 2 perimeter road edges. In the summer of 1990, the site was roller-chopped, which spread cogongrass rhizomes from the road edges into the plantation, so that cogongrass occupied -- 2 ha in a corner of the site. The following winter, bare-root 1-yr-old longleaf pine seedlings were planted mechanically and by hand on a 1.8 x 2.4 in-grid (2244 pines/ha) across the entire plantation. Four years after the longleaf pine were planted, I counted surviving seedlings and measured their heights along 90 in-long line transects following planting
rows. Twelve of the "row-transects" were dominated by dense cogongrass and 12 were randomly located nearby in the same plantation without cogongrass.
Two growing-seasons (16 mo) after seed were planted, there were no seedlings
found for 5 of the 6 species (Dalea, Pityopsis, Sorghastrum, Myrica, and Quercus) in any treatment. Only Serenoa had surviving seedlings. Percent survival of Serenoa seedlings (Table 4-2) was significantly lower in cogongrass than sandhill, and there was a significant difference between unburned and recently burned. There was no interaction between vegetation and fire. Survival of Serenoa seedlings was lower in unburned cogongrass than in any other treatment (Figure 4-1). Growth (shoot and root dry weight) showed no significant differences due to vegetation, fire, or interactions between vegetation and fire (Table 4-3, Figure 4-1).
Table 4-2. ANOVA of percent survival of seedlings of Serenoa repens 2 growingseasons after being planted as seed. Asterisk denotes significant differences (P < 0.05). Source of Variation df F P power
vegetation 1 5.50 0.02* 0.54
fire 1 5.50 0.02* 0.54
vegetation x fire 1 2.63 0.11 0.23
In the seed plots there were no burrowing animal mounds which could be analyzed as a covariate of seedling recruitment.
E shoot a
E 04 L i02a a a
unburned unburned recently- recentlycogongrass sandhill burned burned
Figure 4-1. Percent survival and root and shoot dry weight (after 2 growing seasons) of Serenoa repens seedlings planted as seed in unburned and recently burned cogongrass and sandhill. Error bars represent 1 standard deviation. Different letters represent significant differences (P < 0.05).
C) 0 0
CA ON N It
4 C14 m ON
rA 0 Q 0 0 0
C14 0 0
cn C14 r- oll 4- m
'U : CD 0
0 W 0 C) C) C)
(n CA 00
0 c; c; C5
C% 0 o
C4 1 C'S
ed 0 0 .0
Two growing-seasons (16 mo) after seedlings were planted, there were at least a few survivors of all 6 species in all treatments. Survival in cogongrass was significantly lower than in sandhill for each species (Table 4-4, Figure 4-2). There was no significant fire effect except for Sorghastrum. There was significant interaction between vegetation and fire only for Sorghastrum and Myrica. Only unburned cogongrass differed from the 2 sandhill treatments for Pityopsis, and only the unburned sandhill differed from the 2 cogongrass treatments for Myrica (Figure 4-2). Survival of Dalea (no recently burned treatments) was lower in unburned cogongrass than in unburned sandhill (T= 40.0, P =
Shoot dry weight (growth) was significantly lower in cogongrass than in sandhill for each species (Table 4-5). Growth was significantly lower in unburned compared to recently burned treatments for all species except Myrica. Only Pityopsis showed significant interaction between vegetation and fire. There was no fire effect in cogongrass for Sorghastrum or Myrica, and there was no fire effect in sandhill for Pityopsis, Myrica, or Quercus (Figure 4-3). Quercus growth in unburned cogongrass was lower than in any other treatment. Growth of Dalea (no recently burned treatments) was lower in unburned cogongrass than in unburned sandhill (t = -8.60, P < 0.01, power = 1.00).
In order to compare recruitment success among plant life-forms (regardless of fire effects), I compared survival and growth among all 5 species, burned and unburned treatment data combined. Percent survival (Figure 4-4) in cogongrass was lowest for Dalea, Sorghastrum, and Quercus, which were significantly lower than survival of Pityopsis and Myrica (H = 32.7, P < 0.01). Survival did not differ between any species in
00 r" W)
Cd C14 W) W)
wl, 0 110
.1.) 0 t- C)
00 m C14 eq
0 to V V 64 1
> 0 C) 0 0
0 0 c) o c)
> \0 CN (7*1
rg Cd ci 0
rA Cd to
C Cd 4
0 0 0
0 0 0 0
172ALUnS % U73OLU
CA 0 00
c) 0 (D
0 0 Clq
4) 0 00 kn
5 kn I.,0-0
-, 4) C;
CA q = 0 0
C C 00
C) p O 0
C) C) C)
p p g C> C) C)
0 00 CD C:, c) C)
0 v v v
01 > tr >
(U Cd Cd "0
0 E E
:3 0 0
0 U rA
0 MI M
41 Cd ci
V Cll 0
R, .2: .00
+- cd CIS
0 M 0 "a
r cd 'A C's 0.
C) C> C)
(9) lqs!;)m fap looqs ugaw
b b ab
a a a
Dalea pinnata Pityopsis Sorghastrum Myrica Quercus
graminifolia secundum cerifera geminata
sandhill species Figure 4-4. Mean % survival (after 2 growing seasons) of seedlings of 5 sandhill species planted in cogongrass and in sandhill, data for unburned and recently burned treatments combined. Error bars represent I standard deviation. Different letters represent significant differences (P < 0.05).
sandhill (H = 3.79, P < 0.44). Mean shoot dry weight (Figure 4-5) in cogongrass was significantly higher for Myrica than for the 4 other species, which did not differ from each other (H = 52.8, P < 0.01). In sandhill, growth of Myrica, Pityopsis, and Dalea was significantly > Sorghastrum which was > Quercus (H = 88.0, P < 0.01).
Percent survival of Pinus seedlings was significantly lower in cogongrass (25 10.3) than in sandhill (53 12.0) ( = -6.20, df= 22, P < 0.01, power = 1.00). Mean height of surviving pines was lower in cogongrass (0.05 0.025 m) than in sandhill (0.20 0.072 m) (T = 78.5, P< 0.01). Forty-five percent of the 139 surviving pines in cogongrass were still in the grass stage, compared to only 10% of the 300 surviving pines in sandhill (Figure 4-6). The tallest pine in the cogongrass was only 0.31 m in height, while the tallest pine in sandhill reached 1.60 m.
The seedling plots had very few burrowing animal mounds; 3 older gopher tortoise mounds in sandhill (2 in unburned and 1 in recently burned,) and none in cogongrass. None of the seedlings planted in these mounds (1 Sorghastrum, 1 Quercus, 1 Dalea) survived. Due to small sample sizes per treatment, mounds could not be analyzed as a covariate of seedling recruitment.
Recruitment, measured as survival and growth of 7 plant species representing various sandhill plant life-forms, was significantly lower in cogongrass than in sandhill. The ability of Serenoa to establish from planted seed in dense cogongrass may be due to its large seed with a large endosperm that provides food for early seedling growth, enhancing its ability to compete with cogongrass during its first year of growth. Studies have shown that seedlings emerging from beneath a dense litter mat must allocate more
E= 0.0 -----Dalea pinnata Pityopsis Sorghastrum Myrica Quercus
graminifolia secundum cerifera geminata
Figure 4-5. Standardized shoot dry mass data (after 2 growing seasons) of seedlings of 5 sandhill species planted in cogongrass and in sandhill, data for unburned and recently burned treatments combined. Error bars represent I standard deviation. Different letters represent significant differences (P < 0.05).
55 50 45
40 iE cogongrass (n = 139)
E sandhill (n = 300) 35
30 25 20 15 10
0 0 0 0 0 0
pine height class (m) Figure 4-6. Distribution by height class of 4 yr-old planted longleaf pines in cogongrass and in sandhill.
energy to penetrating it, and that larger-seeded species tend to have higher survival than smaller-seeded species (Facelli and Pickett 1991).
In cogongrass, survival of clonal species was higher than for non-clonal species. In cogongrass, the greatest shoot growth was achieved by Myrica, a clonal woody shrub that is able to fix atmospheric N, another likely advantage when competing with cogongrass for limited soil nutrient resources. This suggests that, in cogongrass swards, clonal growth may enable seedlings to exploit patchily distributed resources, as cogongrass does through rhizomatous growth. A study of invasion in Hawaiian forests showed that Psidium cattleianum (strawberry guava), unlike most indigenous species, produced clonal stems with a higher leaf area than its seedlings, perhaps facilitating dominance of this species in the forest understory (Huenneke 1990).
Burning before planting enhanced seedling recruitment of some sandhill species in both cogongrass and sandhill. In cogongrass, recent burning increased survival and/or growth of all species except Myrica. It may be that the high growth rate of Myrica allows it to successfully compete with cogongrass even in unburned areas.
It follows from the above summary of results, and from the Chapter 3 studies of fire in cogongrass, that successful seedling recruitment in cogongrass that has invaded Florida sandhill is most likely for large-seeded, clonal, N-fixers that are dispersed soon after burning and are highly fire-tolerant. No indigenous sandhill species fits all of these criteria. In addition, successful recruitment does not necessarily lead to permanent establishment, and established seedlings must be able to successfully compete with cogongrass (discussed in Chapter 5) and survive fire in cogongrass as they grow.
Successful seedling recruitment in cogongrass has been shown in studies in
Southeast Asia, which documented plant species growing in association with cogongrass under various conditions (Table 4-6). When cogongrass was burned, most associates were fire-tolerant grasses, ranging from rhizomatous or stoloniferous to caespitose, and from short (< 1 m) to tall (2 1 m). When fire was prevented in cogongrass, succession proceeded to non-pyrogenic forest (Sajise 1972, Eussen and Wirjahardja 1973). In these cases, cogongrass was initially displaced by species such as Eupatorium odoratum, Melastoma affine, and Saccharum spontaneum that were fire-resistant and had higher canopies than cogongrass (Eussen and Wirjahardja 1973). These species also had deeper root systems than cogongrass, which is known to promote niche separation and coexistence between species (Berendse 1981, 1982). When heavily grazed in Malaysia, cogongrass was displaced by Paspalum conjugatum, a stoloniferous grass (Bor 1960). These reports suggest that seedling recruitment in cogongrass in Florida sandhill might be facilitated by preventing fire (if that is feasible). This would likely result in recruitment of non-pyrogenic hardwoods, as occurs in now in fire-excluded sandhill.
Finally, the highly significant decrease in pine seedling survival and height growth in cogongrass is an important consideration for foresters trying to establish longleaf pine where cogongrass is present. In such cases, site preparation procedures that spread cogongrass rhizomes, such as roller-chopping, should be reconsidered or only used in conjunction with cogongrass control, which may be economically justified by the potential for increase in longleaf pine survival and growth. More importantly, in the absence of significant longleaf pine regeneration, sandhill would lose its dominant overstory species and the animals that depend on it, such as red-cockaded woodpecker.
Table 4-6. Plants associated with cogongrass in Southeast Asia (Hubbard et al. 1944, Bor 1960, Eussen and Widahardja 1973).
Agrostis olivetorum Juncus acutus
Agropyron multiflorum Juncus maritimus
A ndropogon spCynodon dactylon Asteraceae:
Eleusine indica Inuld viscosa
Erianthus ravennae Eupatorium odoratum
Panicum maximum Fabaceae:
Panicum reopens Mimosa sp.
Pennisetum sp. Melastomataceae:
Phragmites sp. Melastoma affine
Saccharum munja (= S. bengalense)
Saccharum spontaneum Plantaginaceae:
Aemeda arundinaceae Plantago crassifolia
Cogongrass reduces seedling recruitment in sandhill, potentially reducing the
persistence of pyrogenic grass, pines, and a diversity of other sexually reproducing sandhill
species. In terms of seedling recruitment, cogongrass is not equivalent to sandhill grasses.
COGONGRASS INVASION LIMITS RESOURCES AVAILABLE TO SANDHILL SEEDLINGS
As discussed in the preceding chapter, seedling recruitment is often the critical life history stage for population persistence (Harper 1977, Goldberg 1982, Gross and Werner 1982, Peart 1989). Data presented in that chapter showed that survival and growth of seedlings of indigenous sandhill ("sandhill seedlings") were reduced where cogongrass invaded and dominated sandhill. Without seedling recruitment to replace sandhill vegetation that was displaced by cogongrass invasion (seen in Chapter 2), cogongrass will presumably continue to dominate sandhill.
In this chapter I examine possible causes of reduced sandhill seedling recruitment in cogongrass by studying limitations of resource availability in sandhill where cogongrass dominates. These findings will assist in determining whether cogongrass is a functional equivalent to sandhill understory vegetation in terms of limitation of resources available to seedlings of indigenous sandhill species. I compare several ecological traits for evidence of equivalence. Light, soil water, and soil nutrients are measured to determine if levels of these essential resources are reduced in cogongrass. Next I analyze traits that contribute to resource acquisition: litter depth, leaf area, and rhizome and root depth, mass, and length. Then I assess the relative strength of above- and below-ground competition in cogongrass and sandhill relative to seedling recruitment of a woody and an herbaceous sandhill
species. To evaluate possible impediments to seedling recruitment other than resource limitations, I compare seedling herbivory in cogongrass and sandhill, and discuss reports of cogongrass allelopathy.
Plant community diversity and patterns of dominance are strongly influenced by availability of light, water and nutrient resources. Resource availability is altered by exploitation competition, where an individual reduces limiting resources available to neighbors; by interference competition such as release of allelopathic compounds, where an individual directly harms neighbors; and by apparent competition such as caused by herbivory, where selective feeding by natural enemies reduces an individual's growth or survival relative to neighbors (Connell 1990).
Viewed mechanistically (Goldberg 1990), competition has also been divided into the effect of each species on the abundance of an intermediary and its response to change in the abundance of the intermediary. From this perspective, competitive interactions between plants are viewed through some intermediary such as light, nutrients, pollinators, herbivores or microorganisms. In other words, individual plants can be good competitors either by depleting a resource or by being able to continue growth at depleted resource levels.
In attempting to develop a predictive theory for plant competition, Tilman
proposed that in a mixture of species limited by the same resource, the species with the lowest requirement for that resource will dominate because it depletes the supply of the resource until species with a higher requirement can no longer survive (Tilman 1990). In
this study I examine the ability of cogongrass to deplete resources in Florida sandhill, and assess the ability of sandhill seedlings to grow at depleted resource levels in cogongrass.
Plants may increase resource uptake by allocation to resource-acquiring organs and by morphological alteration of resource-acquiring systems (Goldberg 1990). Morphological features that contribute to the competitive ability of plants include leaf area, vertical distribution of leaf area, root length, and root mass (Berendse and Elberse 1990). To compare the ability of cogongrass and of indigenous sandhill understory vegetation to acquire resources, I compare leaf area, and root depth, mass, and length.
Successful recruitment of seedlings into established vegetation involves
overcoming both above- and below-ground resource limitations exacerbated by size asymmetry (Grubb 1977). For a seedling to germinate, establish, and grow in dense vegetation, it must overcome limitations of light, space, soil water, and soil nutrients. The relative importance of these various resources depends on plant traits such as growth form, physiology, and symbiotic relationships. To determine if sandhill seedlings are limited by light or soil resources, I compare above- and below-ground resource limitations to recruitment in cogongrass and sandhill of 2 species indigenous to sandhill, one woody and the other herbaceous.
In studies of interference competition, separating effects due to allelopathy from those due to exploitation competition or apparent competition (herbivory) is difficult (Putnam and Tang 1986). Although I did not specifically test for cogongrass allelopathy in this study, at the end of this chapter I discuss published cogongrass allelopathy studies in order to review evidence for cogongrass allelopathy in Florida sandhill.