Using Gps/Gis and Molecular Analysis As Conservation Management Tools for Georgia Plume

CONSERVING OUR BOTANICAL HERITAGE: USING GPS/GIS AND MOLECULAR

ANALYSIS AS CONSERVATION MANAGEMENT TOOLS FOR GEORGIA PLUME

JUSTIN PORTER

(Under the Direction of David Berle and Hazel Y. Wetzstein)

ABSTRACT

Georgia plume (Elliottia racemosa) is a threatened tree endemic to Georgia. Current records regarding the species are difficult to use and only very general habitat descriptions are available.

This study developed a GIS based conservation management tool using Georgia plume as a model system. Field surveys collected census information and habitat data for several populations using putative sightings provided by the Department of Natural Resources. An alarming reduction in the number of populations has occurred. Approximately 40% of visited populations no longer contained Georgia plume, with 58% of population loss attributed to human activity. Random Amplified Polymorphic DNA was used to evaluate genetic relationships among and within selected Georgia plume populations. Genetic analysis revealed that populations containing fewer individuals tended to have lower genetic diversity while populations with larger numbers tended to have higher genetic diversity. Two populations with numerous individuals that were sampled extensively displayed the greatest genetic diversity.

INDEX WORDS: Georgia plume, Conservation, Geographic Information System (GIS),

Random Amplified Polymorphic DNA (RAPD)

CONSERVING OUR BOTANICAL HERITAGE: USING GPS/GIS AND MOLECULAR

ANALYSIS AS CONSERVATION MANAGEMENT TOOLS FOR GEORGIA PLUME

(ELLIOTTIA RACEMOSA).

JUSTIN PORTER

B.S.A., The University of Georgia, 2007

A Thesis Submitted to the Graduate Faculty of the University of Georgia in Partial Fulfillment of

the Requirements for the Degree

MASTER OF SCIENCE

ATHENS, GEORGIA

2010

JUSTIN PORTER

CONSERVING OUR BOTANICAL HERITAGE: USING GPS/GIS AND MOLECULAR

ANALYSIS AS CONSERVATION MANAGEMENT TOOLS FOR GEORGIA PLUME

(ELLIOTTIA RACEMOSA).

JUSTIN PORTER

Major Professors: David Berle Hazel Y. Wetzstein

Committee: Elizabeth Kramer Robert Trigiano

Maureen Grasso Dean of the Graduate School The University of Georgia July 2010

ACKNOWLEDGMENTS

I would like to thank everyone who has helped me along the way with this project. Lisa

Kruse and Jennifer Ceska for helping initially develop the project. Lara Jackson for braving the thickest woods we’ve ever seen. Denita Hadziabdic for a roof in Knoxville and Phillip Wadl for helping me in the lab there. Henry Mincy for driving around Fort Stewart and showing us the trees. Kevin Samples for helping me with GIS.

Finally thanks to my family and friends for all the encouragement, support, laughs, trivia, games, fall Saturdays, trips, weddings, parties, movies, etc. How sweet it is…Go Dawgs.

TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ...... iv

LIST OF FIGURES ...... vii

LIST OF TABLES...... viii

CHAPTER

1 INTRODUCTION AND LITERATURE REVIEW ...... 1

1.1 PLANT DESCRIPTION AND HISTORY...... 1

1.2 GEOGRAPHIC INFORMATION SYSTEMS IN

CONSERVATION APPLICATIONS...... 4

1.3 RANDOM AMPLIFIED POLYMORPHIC DNA ...... 8

1.4 OBJECTIVES AND APPROACH...... 9

1.5 CONCLUSION...... 10

REFERENCES ...... 11

2 USING GIS/GPS AS A CONSERVATION MANAGEMENT TOOL: GEORGIA

PLUME AS A CASE STUDY ...... 17

2.1 INTRODUCTION ...... 19

2.2 MATERIALS AND METHODS...... 22

2.3 RESULTS AND DISCUSSION...... 24

2.4 CONCLUSION...... 32

REFERENCES ...... 33

3 MOLECULAR ANALYSIS OF THE THREATENED ENDEMIC GEORGIA PLUME

USING RANDOM AMPLIFIED POLYMORPHIC DNA...... 46

3.1 INTROCUCTION...... 48

3.2 MATERIALS AND METHODS...... 51

3.3 RESULTS AND DISCUSSION...... 52

REFERENCES ...... 58

4 CONCLUSIONS...... 70

vii

LIST OF FIGURES

Page

Figure 1.1. Georgia plume in flower...... 15

Figure 1.2. The distribution of previously reported Georgia plume populations in Georgia...... 16

Figure 2.1. Counties proposed to contain Georgia plume populations...... 37

Figure 2.2. Data entry form developed for use in the field with ArcPad...... 38

Figure 2.3. Map showing Georgia counties and locations of putative Georgia plume population sites from DNR data base. Dotted triangles are sites that were visited for data collection while dots show putative locations. Labels indicate EO numbers of respective populations ...... 39

Figure 2.4. Detail of habitat suitability model. Darker areas correspond to more favorable habitat based on overlay analysis while white areas correspond to unfavorable areas for Georgia plume...... 40

Figure 3.1. Georgia plume in flower...... 61

Figure 3.2. Georgia plume fruit capsules...... 62

Figure 3.3. Populations sampled for DNA analysis...... 63

Figure 3.4. The location of individual plants sampled within the Charles Harrold Preserve population ...... 64

Figure 3.5. The location of individual plants sampled within the Big Hammock Wildlife Management Area population...... 65

Figure 3.6. Dendrogram of genetic relationships using Jacard’s coefficient of selected populations. Different letters reflect different populations while numbers reflect individual samples within respective populations. Colored bars show samples from respective populations...... 66

Figure 3.7. Dendrogram of genetic relationships using Jacard’s coefficient of similarity within Charles Harrold Population...... 67

Figure 3.8. Dendrogram of genetic relationships using Jacard’s coefficient of similarity within Big Hammock Population...... 68

viii

LIST OF TABLES

Page

Table 2.1. Sites visited and the number of data points recorded at each to reflect changing environmental conditions at each site...... 41

Table 2.2. Site characteristics of locations where Georgia plume is present as verified by field visits ...... 42

Table 2.3. Characteristics of Georgia plume trees in active population sites...... 43

Table 2.4. Site characteristics of locations where Georgia plume is not present as verified by field visit ...... 44

Table 2.5. Assessment of the occurrence of Georgia plume at unvisited sites based on surveys of aerial photographs...... 45

CHAPTER 1

INTRODUCTION AND LITERATURE REVIEW

1.1 PLANT DESCRIPTION AND HISTORY

Georgia plume (Elliottia racemosa Muhlenberg ex Elliott [Ericaceae]) is a small tree that is endemic only to the State of Georgia, where it is listed as a threatened species (Georgia

Department of Natural Resources 2006). It is known mainly for its large, white, plume-like inflorescences that are born in June and July (Fig. 1.1). It is confined mainly to the Coastal Plain province and occurs in approximately 36 locations (Fig. 1.2), though exact numbers are difficult to determine (Duncan and Duncan 1978; Chafin 2007). There were supposedly populations found in South Carolina, however, these reports have never been substantiated and it is now believed that these were cultivated populations (Fordham 1991).

There is considerable confusion and disagreement as to who was the first to describe

Georgia plume. It is commonly held that William Bartram first discovered the tree along the

Savannah River in the 1770’s. His description, however, was apparently unpublished. The first published description comes from Steven Elliott in 1808. The genus was later named Elliottia in

1810 by Henry Muhlenberg to honor Elliott (Del Tredici 1987).

From approximately 1875 to 1901 it was believed that Georgia plume was extinct in the wild (Chafin 2007). This was though to be partly caused of the lack of seed production that several authors have noted (Faircloth 1970; Miller 1978; Del Tredici 1987; Fordham 1991).

There has also been a lack of seedlings observed in the wild. Low to no fruit production has been observed for much of the plant’s known history (Elliott 1821). Recent work has shown that 2

Georgia plume exhibits gametophytic self-incompatibility (Radcliffe 2009). Low genetic

diversity has been reported from allozyme analysis of several populations (Godt and Hamrick

1999). The combination of these two factors explains the biological reasons for the decline of the

species in the wild. Another major factor leading to its decline is habitat loss. Clearing of forest

land for use as agricultural fields and timber land are major contributors to Georgia plume

habitat loss. The idea that removal of a natural fire regime has disrupted the natural seed

germination ecology has also been put forth (Chafin 2007).

Other Studies on Georgia plume

Propagation

Study of Georgia plume has mainly focused on propagation, the goal being to produce

more plants for safeguarding and for specimens. Those focusing on seeds have encountered some problems such as a lack of seed production in the wild. This has led to ambiguity as to seed

germination requirements. Although plants generally produce large amounts of flowers, few if any fruit are produced and with mature fruit generally producing little seed. Studies have been carried out that involve cold stratification and liquid smoke extract but have been unable to make any definitive conclusions due to the lack of seed available (Fordham 1991). This lack of seed production in the wild has led to the idea that plants are self-incompatible and that populations are made up of relatively few inbred clones (Del Tredici 1987).

Floral Biology

Radcliffe (2009) determined that Georgia plume exhibits sporophytic self-incompatibility

through pollen tube staining. This shows that cross pollination form a genetically distinct

individual is required for sexual reproduction to occur. Pollen viability testing also showed that

pollen viability time was extremely short. These factors combined with the fragmented nature of 3

populations appear to be contributing to low sexual reproduction in general. Future studies in

Georgia plume floral biology will focus on pollinator ecology and will include controlled cross

pollination between individuals with known genetic relationships to quantify fruit and seed

production. Seed viability testing might also be performed on wild-collected seeds.

Tissue Culture

Tissue culture was recently investigated as a conservation approach for Georgia plume

(Woo 2007). The approach was presented as an alternative to traditional propagation methods

because of their inefficiency. A method for sterilization and media preparation was developed

that gave a high percentage of success when producing new plantlets for transplantation. This

study was able to produce rooted plants that were used in current outplanting studies in an

attempt to further define cultural requirements for the plant. Treatments have included

inoculation with mycorrhizae, inclusion of supplemental organic matter, and protection from

deer browsing.

Allozymes

The genetic diversity of Georgia plume has received little attention to date. Only one

study (Godt and Hamrick 1999) has been carried out to study this directly. Allozymes were

utilized to investigate the supposed clonality of populations that was thought to result from

extensive root sprouting. Genetic diversity of the species and populations were investigated. It was determined that genetic diversity at the species level was low when compared to other woody plants and that most of the genetic diversity was present at the population level.

Populations with the highest diversity were those that were larger and were nearer to the center of the range of distribution. In populations where genetic diversity was low, seed set was also

found to be lower. These populations only consisted of a few clones. 4

1.2 GEOGRAPHIC INFORMATION SYSTEMS IN CONSERVATION APPLICATIONS

A Geographic Information Systems (GIS) is a spatially enabled database system that allows the user to integrate and query diverse data sets. It provides several advantages to a traditional paper map. A GIS provides a way to collect, store and organize information, perform analysis on it, and to create visualizations of data (Salem 2003). Combining these functions provides a powerful tool set to conservationists that enable them to make more informed decisions. Use of a geographic approach is becoming common in studies because including spatial information provides more complete or more meaningful data than could otherwise be obtained from basic attribute information.

One powerful GIS tool is the ability to model the distribution of an event in relation to other factors such as environment through the use of spatial statistics. This is an extremely useful conservation tool as many distribution maps are based on large areas that are interpolated from known occurrence sites and may be biased toward well-sampled or easily accessible areas

(Giriraj, Irfan-Ullah et al. 2008). The ability to model the occurrence of an event is made possible by the fact that landscape scale maps of environmental variables are now available

(Bourg, McShea et al. 2005). It is important for conservationists to know where a species is located for efficient conservation measures to be put in place. This information will allow efforts to be more efficiently focused and can examine interactions with other variables that may be present. By employing a GIS and several types of environmental data good habitat can be identified for conservation of a threatened species (Parviainen, Luoto et al. 2008).

GIS in other studies/modeling approaches

A geographic approach can be applied to a wide range of problems. There are a number

of diverse fields that utilize GIS. Notably, plant conservation is a field that has embraced this

technology because of the advantages it offers to traditional mapping and inventory applications.

In a study by Alexander et al. (2005) environmental variables were included in a simple

overlay analysis to determine a possible distribution range for American chestnut (Castanea dentate). Layers in this analysis included satellite imagery, soil type, vegetation types and topography. This information was combined with known locations of trees to roughly predict other possible locations of trees. The aim of this study was to identify unique areas for the collection of plant material to be used in a breeding program. The authors were also able to identify field locations for reintroduction of trees. They also collected material for ex situ

conservation and indexed this material using geographic coordinates.

Vanderpoorten et al. (2005) used partial least squares regression modeling in GIS to

determine what land use and soil variables affected the distribution of bryophytes in the Semois

river basin. Using this approach they were able to determine that steep slopes and military land

were important for the conservation of not only bryophytes but also for biodiversity in general.

The authors suggested that this approach could be used to identify other land for conservation

protection and had the benefit of not being influenced by existing administrative boundaries.

GIS can be used to model the occurrence of invasive species to identify areas for

remediation. Draper et al. (2003) was able to identify which areas specific invasive species were

likely to be present so that land managers could take action to protect native threatened species.

This study used simple landscape variables that included altitude and slope. 6

As can be seen from a brief review of literature dealing with predicting distribution of

species there are several approaches. This form of modeling can be applied to both plants and

animals. It is probably somewhat easier to develop a distribution model for plants due to the fact that they do not move around. The selection of a method depends on the problem at hand.

The Classification and Regression Tree method (CART) is one such method that has been used to identify potential distributions of events based on other variables (Bourg, McShea et al. 2005; Cheng, Zhang et al. 2009; Kumar, Spaulding et al. 2009). This method utilizes points of known occurrence and absence and evaluates the importance of environmental variables at those points. A decision tree is then created that is used in later map development. An advantage of this method is that it reveals the relative importance of each variable in regards to the distribution. CART models were found to be better at predicting species distributions than

General Linear Models (Vayssieres, Plant et al. 2000). Another advantage is that this type of model is that it is better able to show interactions among variables being examined.

Other methods used for modeling species distribution include logistic regression, kriging, spatial regression techniques, and Genetic Algorithm for Rule-set Production (GARP). There are a variety of software packages and extensions that have been developed to implement these methods.

Logistic regression uses simple presence or absence of the process being examined as the dependent variable and is able to estimate the dependent variable in unknown locations based on other variables values. A portion of study points are used as training points to examine study variables in relation to the dependent variable while the remaining portion of study points are used as validation points to check the accuracy of the resulting regression function. This method 7 has been used in several studies to assess possible habitat distribution (Kumar, Spaulding et al.

2009) and to assess habitat loss as a result of development (Pereira and Itami 1991).

Kriging is a form of spatial interpolation that is useful for producing estimation surfaces based on known survey points. It is based on the idea that near points are more similar to each other than farther points. Unlike classical statistical techniques that assume samples are independent from each other, kriging assumes that spatial dependence between points does exist.

This allows attribute values at unknown locations to be estimated using a predetermined weighting scheme. Traditional kriging is only able to estimate values for one variable. More advanced forms of kriging are able to integrate several variables.

There are several types of spatial regression techniques that have been developed. These include the spatial lag model, spatial error model, and Geographically Weighted Regression

(GWR). Each of these models makes slightly different assumptions and may be better at explaining relationships among variables than another model based on the nature of the dependent variable. In the spatial lag model the dependent variable is considered to be autoregressive across space while the spatial error model assumes that only the error term is autocorrelated. Using GWR spatial dependence is implicitly accounted for but the model’s main goal is to account for spatial heterogeneity. When using these models it is often best to run each model and use a variety of diagnostic tests to choose the one that best explains the relationships among the independent and dependent variables.

The GARP method uses several different algorithms in a iterative manner to develop rules that define a particular species niche (Peterson 2001). These rules are based on areas oof known occurrence and conditions at those sites and are then used to interpolate to other areas. It 8 is useful is determining a species niche. GARP is often considered a superset of other modeling techniques because it integrates several different procedures.

1.3 RANDOM AMPLIFIED POLYMORPHIC DNA

Evaluating the genetic relationships of a species is an important part of a conservation effort. This information gives valuable insight into the genetic structure of populations and could play a role in protecting the genetic makeup of a species but also in developing breeding programs for the conservation of diversity.

There are several methods available that can be used to determine these relationships.

Random Amplified Polymorphic DNA (RAPD) is one such method that is relatively simple and inexpensive. RAPDs work by using an arbitrary primer to amplify portions of sample DNA during polymerase chain reaction. Polymorphisms are detected among samples when portions of

DNA with differing sequence amplify differently. A major benefit is that no knowledge of actual sequence is required. Amplification products are then separated on agarose gels for visualization and rating. The resulting image is a unique banding pattern for each sample that, when compared to other samples over several primers, can be used to determine similarity to other samples.

Several other recent studies have used RAPDs to assess genetic diversity, investigate population genetic structure, or to determine parentage (Csergo, Schonswetter et al. 2009;

Gonzalez-Perez, Sosa et al. 2009; Subramanyam, Muralidhararao et al. 2009; Vyas, Sharma et al.

2009). RAPDs can determine relationships among individuals (Neuhaus, Kuhl et al. 1993;

Gunter, Tuskan et al. 1996) and thus is a method that can provide important information in plant conservation programs. Gonzalez-Perez et al. (2009) used RAPDs investigated genetic diversity in the endangered legume Anagyris latifolia that is endemic to the Canary Islands. This species consists of small, isolated populations similar to Georgia plume. Schiebold et al. (2009) assessed 9

genetic diversity in the German rare endemic grass Calamagrostis pseudopurpurea and was able to identify clones of the same genotype. Kandedmir et al. (2004) investigated genetic variation in different populations of Turkish pine (Pinus brutia). Information from this study was used to understand genetic variation for use in breeding programs and to help conserve genetic resources of this economically important tree species. Hollman et al. (2005) used this technique to identify putative clones of velvet bent grass.

Although there are concerns about the reproducibility of RAPD markers (Devos and Gale

1992) there are several advantages to this method. Chief among these are the speed and ease with which they can be carried out and their low cost when compared to other methods such as

Amplification Fragment Length Polymorphism (AFLP), Simple Sequence Repeat (SSR), DNA

Amplification Fingerprinting (DAF), or microsatellites. Penner et al. (1993) found that

reproducibility of RAPD amplification products could be greatly increased when protocols were diligently followed and that dissimilar results were caused by factors such as thermocycler programs that were out of calibration.

1.4 OBJECTIVES AND APPROACH:

There were two main objectives of this study. The first was to characterize the environmental requirements for Georgia plume and produce a more accurate habitat distribution.

This was accomplished by integrating georeferenced environmental data on a landscape scale and field observations with known locations of Georgia plume. Raster overlay was used to find areas that contained favorable environmental variables were Georgia plume was present and to interpolate other areas where it might occur.

The second was to determine the genetic relationships within and among several different populations to answer questions about clonality. Populations were defined as a group of 10

individuals that had a continuous distribution over a specific area (Li X. J., Yang et al. 2008).

This was done by performing a modified RAPD analysis on individuals from several populations

and developing a similarity matrix along with a dendrogram and comparing these with physical

distances.

1.5 CONCLUSION

The methods presented here are very useful in a wide variety of conservation projects.

They could be easily adapted for use on other endangered and threatened plant species in future projects. Increasing availability of GIS data along with continually decreasing technology process are making GIS analysis more feasible and more accurate. At the same time the ease and low cost of RAPD analysis, when compared to other methods, make it an ideal method for use in a wide variety of conservation programs. 11

REFERENCES

Alexander, M. T., J. H. Craddock, et al. (2005). "Conservation of Castanea dentata germplasm

of the Southeastern United States." Acta Horticulturae: 485-490.

Bourg, N. A., W. J. McShea, et al. (2005). "Putting a cart before the search: Successful habitat

prediction for a rare forest herb." Ecology 86(10): 2793-2804.

Chafin, L. J. (2007). Field guide to the rare plants of Georgia. Athens, GA, The State Botanical

Garden of Georgia.

Cheng, W., X. Y. Zhang, et al. (2009). "Integrating classification and regression tree (CART)

with GIS for assessment of heavy metals pollution." Environmental Monitoring and

Assessment 158(1-4): 419-431.

Csergo, A. M., P. Schonswetter, et al. (2009). "Genetic structure of peripheral, island-like

populations: a case study of Saponaria bellidifolia Sm. (Caryophyllaceae) from the

Southeastern Carpathians." Plant Systematics and Evolution 278(1-2): 33-41.

Del Tredici, P. (1987). "Lost and Found: Elliottia racemosa." Arnoldia 47(4): 3-8.

Devos, K. M. and M. D. Gale (1992). "The use of Random Amplified Polymorphic DNA

Markers in Wheat." Theoretical and Applied Genetics 84(5-6): 567-572.

Draper, D., A. Rossello-Graell, et al. (2003). "Application of GIS in plant conservation

programmes in Portugal." Biological Conservation 113(3): 337-349.

Duncan, W. H. and M. B. Duncan (1978). "Elliottia in the Piedmont of Georgia." Castanea

43(2): 182-184.

Elliott, S. (1821). A Sketch of the Botany of South Carolina and Georgia. Charleston, SC, J.R.

Schenck. 12

Faircloth, W. R. (1970). "An Occurence of Elliottia in Central South Georgia." Castanea 35(1):

58-61.

Fordham, A. J. (1991). "Elliottia racemosa and its propagation." Arnoldia 51(4): 59-62.

Georgia Department of Natural Resources (2006). Georgia's State Protected Animals and Plants.

Wildlife Resources Division - Nongame Conservation Section. Social Circle, GA.

Giriraj, A., M. Irfan-Ullah, et al. (2008). "Mapping the potential distribution of Rhododendron

arboreum Sm. ssp nilagiricum (Zenker) Tagg (Ericaceae), an endemic plant using

ecological niche modelling." Current Science 94(12): 1605-1612.

Godt, M. J. W. and J. L. Hamrick (1999). "Population genetic analysis of Elliottia racemosa

(Ericaceae), a rare Georgia shrub." Molecular ecology 8: 75-82.

Gonzalez-Perez, M. A., P. A. Sosa, et al. (2009). "Genetic variation and conservation of the

endangered endemic Anagyris latifolia Brouss. ex Willd. (Leguminosae) from the Canary

Islands." Plant Systematics and Evolution 279(1-4): 59-68.

Gunter, L. E., G. A. Tuskan, et al. (1996). "Diversity among populations of switchgrass based on

RAPD markers." Crop Science 36(4): 1017-1022.

Hollman, A. B., J. C. Stier, et al. (2005). "Identification of putative velvet bentgrass clones using

RAPD markers." Crop Science 45(3): 923-930.

Kandedmir, G. E., I. Kandemir, et al. (2004). "Genetic variation in Turkish red pine (Pinus brutia

Ten.) seed stands as determined by RAPD markers." Silvae Genetica 53(4): 169-175.

Kumar, S., S. A. Spaulding, et al. (2009). "Potential habitat distribution for the freshwater diatom

Didymosphenia geminata in the continental US." Frontiers in Ecology and the

Environment 7(8): 415-420. 13

Li X. J., H. L. Yang, et al. (2008). "Genetic variation within and between populations of the

Qinghai-Tibetan Plateau endemic Gentiana straminea (Gentianaceae) revealed by RAPD

markers." Belgian Journal of Botany 141(1): 95-102.

Miller, H. A. (1978). "The story of Elliottia, a primitive, slow-growing member of the heath

family that is rare in more ways than one." American Forests 84(2).

Neuhaus, D., H. Kuhl, et al. (1993). "Investigation on the Genetic Diversity of Phragmites

Stands using Genomic Fingerprinting." Aquatic Botany 45(4): 357-364.

Parviainen, M., M. Luoto, et al. (2008). "Modelling the occurrence of threatened plant species in

taiga landscapes: methodological and ecological perspectives." Journal of Biogeography

35(10): 1888-1905.

Penner, G. A., A. Bush, et al. (1993). "Reproducibility of Random Amplified Polymorphic DNA

(RAPD) Analysis among Laboratories." Genome Research 2: 341-345.

Pereira, J. M. C. and R. M. Itami (1991). "GIS-Based Habitat Modeling using Logistic Multiple-

Regression - A Study of the MT. Graham Red Squirrel." Photogrammetric Engineering

and Remote Sensing 57(11): 1475-1486.

Peterson, A. T. (2001). "Predicting species' geographic distributions based on ecological niche

modeling." Condor 103(3): 599-605.

Radcliffe, C. A. (2009). Studies in reproductive biology and in vitro propagation as approaches

for the conservation of Elliottia racemosa. Department of Horticulture. Athens, The

University of Georgia. M.S.

Salem, B. B. (2003). "Application of GIS to biodiversity monitoring." Journal of Arid

Environments 54(1): 91-114. 14

Schiebold, S., I. Hensen, et al. (2009). "Extensive clonality of the endemic Calamagrostis

pseudopurpurea Gerstl. ex OR Heine in central Germany revealed by RAPD markers."

Plant Biology 11(3): 473-482.

Subramanyam, K., D. Muralidhararao, et al. (2009). "Genetic diversity assessment of wild and

cultivated varieties of Jatropha curcas (L.) in India by RAPD analysis." African Journal

of Biotechnology 8(9): 1900-1910.

Vanderpoorten, A., A. Sotiaux, et al. (2005). "A GIS-based survey for the conservation of

bryophytes at the landscape scale." Biological Conservation 121(2): 189-194.

Vayssieres, M. P., R. E. Plant, et al. (2000). "Classification trees: An alternative non-parametric

approach for predicting species distributions." Journal of Vegetation Science 11(5): 679-

694.

Vyas, G. K., R. Sharma, et al. (2009). "Diversity analysis of Capparis decidua (Forssk.) Edgew.

using biochemical and molecular parameters." Genetic Resources and Crop Evolution

56(7): 905-911.

Woo, S. M. (2007). In Vitro Propagation of Georgia plume, Elliottia racemosa, a Threatened

Georgia Endemic, University of Georgia.

Figure 1.1. Georgia plume in flower. 16

Figure 1.2. The distribution of previously reported Georgia plume populations in Georgia. 17

CHAPTER 2

USING GIS/GPS AS A CONSERVATION MANAGEMENT TOOL: GEORGIA PLUME

AS A CASE STUDY1

1 Justin Porter, David Berle, Hazel Y. Wetzstein, and Elizabeth Kramer. To be submitted to Southeastern Naturalist

ABSTRACT

Georgia plume is a threatened plant endemic only to Georgia. Current records regarding the

species are difficult to use and only a very general habitat description is available from several different sources. The goal of this study was to develop a GIS based conservation management tool using Georgia plume as a model system. A second goal was to develop a habitat suitability model for the species using existing environmental geodata. Field surveys were conducted to collect habitat data for several populations of Georgia plume using putative sightings provided by the Department of Natural Resources. Raster analysis was used to develop a habitat suitability model. Approximately 40% of visited populations were found to no longer contain Georgia plume and of these 70% were found to be due to human activity. Overall habitat conditions appeared to be highly variable with no single factor seeming to explain the presence of the species. The habitat suitability model identified approximately 34% of the study area as being

either highly or moderately suitable for Georgia plume. These areas could be useful for future

conservation areas or reintroduction programs.

Keywords Georgia plume, Geographic Information System (GIS), Habitat Survey, Conservation

Management, Habitat Suitability Model

2.1 INTRODUCTION

Georgia plume (Elliottia racemosa Muhlenberg ex Elliott [Ericaceae]) is a small tree that

is endemic only to the State of Georgia, where it is listed as a threatened species (Georgia

Department of Natural Resources 2006). It is known mainly for its large, white, plume-like

inflorescences that are born in June and July. It is confined mainly to the Coastal Plain region

and occurs in approximately 36 locations, though exact numbers are difficult to determine

(Duncan and Duncan 1978). There were supposedly populations found in South Carolina,

however, these reports have never been substantiated and it is now believed that these were

cultivated populations (Fordham 1991).

There is considerable confusion and disagreement as to who was the first to describe

Georgia plume. It is commonly held that William Bartram first discovered the tree along the

Savannah River in the 1770’s. His description, however, was apparently unpublished. The first

published description comes from Steven Elliott in 1808. The genus was later named Elliottia in

1810 by Henry Muhlenberg to honor Elliott (Del Tredici 1987).

From approximately 1875 to 1901 Georgia plume was thought to be extinct in the wild

(Chafin 2007). The extreme rarity of the plant is caused in part by lack of or limited seed

production that several authors have noted (Faircloth 1970; Miller 1978; Del Tredici 1987;

Fordham 1991). In recent times, seedling recruitment has not been observed in the wild. Low to

no fruit production has been observed for much of the plant’s known history (Elliott 1821).

Recent work has shown that Georgia plume exhibits gametophytic self-incompatibility and has very short pollen viability time (Radcliffe 2009). Low genetic diversity has been reported from allozyme analysis of several populations (Godt and Hamrick 1999). The combination of these factors explains the biological reasons for the decline of the species in the wild. Another major

factor leading to its decline is habitat loss. Clearing of forest land for use as agricultural fields

(Duncan and Duncan 1978) and timber land are major contributors to Georgia plume habitat

loss. The idea that removal of a natural fire regime has disrupted the natural seed germination ecology has also been put forth (Chafin 2007).

The apparently wide variety of habitats but limited number of locations in which Georgia plume occurs has made it difficult to draw any definitive conclusions as to the distribution of the

species. Current distribution maps are in disagreement as to where populations do occur, are

somewhat general, and are based on plant sightings rather than environmental requirements of

the plant.

Currently, records concerning Georgia plume populations are scattered and difficult to

use. The use of paper maps makes updating and sharing records exceedingly difficult. While

some Georgia plume populations have been recorded with the Global Positioning System (GPS),

the majority of populations are only approximately located based on general descriptions of

location. A Geographic Information System (GIS) offers several advantages over the use of

paper maps and paper record keeping. By integrating GPS into the system errors made in

approximating site locations to paper maps can be removed and exact locations can be recorded.

This is paramount when locating Georgia plume populations as they tend to be small in size.

Returning researchers will have a much easier time finding and returning to a population when exact locations are determined. Furthermore, current GPS technology has the accuracy to record the location of individual trees within populations. This makes it possible to conduct censuses of populations and determine distances between individuals for use in other studies, such as genetic analysis.

Another advantage of keeping population records in a GIS is the ability to store several

types of attributes easily. Paper versions of charts can be unwieldy to use, difficult to interpret,

and be subject to introducing errors. GIS offers flexibility in recording information as needed.

Attribute information can be deleted and new attributes can be recorded providing instantly

updated data sets. Also by recording attributes concurrently with spatial information, the

possibility for errors in data entry is reduced. Since data are entered in the field digitally there is

no need to re-enter it later and possibly introduce more errors. Finally because attributes are

stored with locations of populations or individuals, analyses can be conducted on these attributes

in relation to other environmental variables. This would lead to a better understanding of the diverse environmental conditions in which Georgia plume occurs.

Several different types of conservation projects have begun adopting the use of geographic approaches (Anderson and Martinez-Meyer 2004; Alexander, Craddock et al. 2005;

Vanderpoorten, Sotiaux et al. 2005). GIS allows analysis to be conducted across several different layers of environmental variables (Clark, Dunn et al. 1993). This enables models to be developed that characterize current habitats and predict unknown habitat locations (Rotenberry, Preston et al. 2006). Knowledge concerning the exact locations and environmental requirements of threatened and endangered plants is crucial when developing management plans. With this information, managers can make informed decisions regarding establishment of protected areas, collection for ex situ conservation, and exploration for unknown populations.

The goal of this study was to 1) develop a GIS based conservation management tool for

Georgia plume to record habitat conditions and inventory populations and 2) to quantitatively

describe the environment where populations exist by developing a habitat suitability model using

existing GIS datasets. The approach was to develop a method to collect, store, and analyze spatial

and environmental data for populations of Georgia plume. While information presented here is

specifically related to Georgia plume, the methodology could readily be use as a model system

for other threatened or endangered species.

2.2 MATERIALS AND METHODS

StudyU Area

The study area included the known distribution of Georgia plume (Fig. 2.1) which is

contained within 16 counties: Ben Hill, Bryan, Bulloch, Burke, Candler, Coffee, Columbia,

Emanuel, Evans, Irwin, Jeff Davis, Long, Tattnall, Telfair, Turner, and Wheeler. It was confined

mainly to the Coastal Plain of eastern Georgia, however one population was located in the lower

Piedmont region. The entirety of this area is located within UTM zone 17N. It is from 3729724

m north to 3471082 m south and 486972 m east to 234671 m west.

U Putative site locations for Georgia plume populations were obtained in shapefile format

from the Georgia Department of Natural Resources. These locations ranged from recently

verified populations to historical annotations. Using these points as a reference, requests were

sent to landowners for permission to visit and document conditions at as many sites as possible.

Permission was obtained to visit 32 populations. Individual populations are referred to by their

Element Occurrence (EO) number.

At sites that were visited, a data collection form was developed for use in ArcPad (ESRI

2008) to record information. Variables assessed included whether Georgia plume plants were

still present at the site, evidence of reproduction (flowering or fruiting), ground conditions,

canopy conditions, and land use (Fig. 2.2). Data were recorded at the approximate center of sites

that had homogenous conditions throughout the population. At larger sites where habitat characteristics changed throughout the population, data were recorded in the disparate areas to

capture the changes. At sites where Georgia plume was present, all individuals were marked

using GPS except in Big Hammock and Fort Stewart where plants were too abundant to record.

Sites that could not be assessed by a field visit due to lack of landowner permission were examined using aerial photographs (ESRI 2008). Land use at a site, such as pine plantations or clear-cuts, were used as the main estimators for absence of Georgia plume plants. Clear cut pine stands and planted pine stands are clearly visible from aerial photography. Based on canopy

cover and apparent site conditions, aerial photographs were used to determine whether the site

appeared to contain Georgia plume.

Habitat Data Collected

Several pieces of environmental data were collected at the visited sites, and included

whether there was evidence of burning suggested by charred bark on trees or burned ground plants. The general habitat type was categorized into types which included flatwoods, sandhills,

oak hammock, mesic, palmetto woods, sandstone outcrops, slopes, mixed bottom lands, and transitions between slope and bottom land. The type of soil cover (ground plants, leaves and needles, or a mixture of the two) was recorded along with the duff (litter layer) thickness on the soil surface. Canopy characteristics such as type (deciduous, evergreen, or mixed), and closure

(open, partially closed, or mostly closed) were recorded. Habitat conditions at sites that did not

contain Georgia plume plants were also recorded. At locations where Georgia plume plants were

found, tree characteristics were recorded for evidence of reproduction, position in the canopy

(understory or overstory), the general form (sucker, shrub, or tree) of the trees present, and the total number of individuals present in a population.

Habitat Suitability Model

A habitat suitability map was developed by using several previously existing raster maps of environmental variables. Raster resolution was 30 meters. Habitat variables included mean fire return interval (USDA Forest Service 2006), land cover, vegetation type, and topographic relative moisture index (NARSAL 2010). These variables were chosen because they contained values that may be important in influencing where Georgia plume occurs. The topographic relative moisture index (TRMI) was used as a surrogate for factors such as slope, aspect, elevation, and slope position. This index integrates these variables into a relative index of soil moisture for unique locations that ranges from 0 to 60 where 60 is wetter.

An extraction function was used to determine the values of each habitat variable at each of the respective locations where Georgia plume was present. These variables were then reclassified in a binary fashion as favorable (raster value = 1) while all other values of habitat variables were considered to be unfavorable (raster value = 0) for Georgia plume. Areas within city limits as well as bodies of water were also considered unsuitable and excluded from analysis. Raster addition was next used to find areas that had high suitability values (these corresponded to areas that had favorable habitat conditions in all or most of the variable layers).

In this way highly suitable locations had a raster value of 4 while somewhat favorable locations had a raster value of 3. Locations were considered unsuitable for Georgia plume if they had a raster value of 2 or less.

2.3 RESULTS AND DISCUSSION

FieldU Visits/ Environmental Data

The data entry form designed for ArcPad worked extremely well in the field and was very flexible. Upon marking the location of a population, the form could be used to enter a

predetermined set of answers or to enter data that were not predicted. A set of general habitat

types was entered in the form but unanticipated types, such as palmetto woods, could be added as

needed in the field.

Sixty one putative Georgia plume populations were identified from records provided

from the Department of Natural Resources. It should be noted that recorded annotations for some

population indicated for some populations may not have been visited for some time and dated

back 31 years. We elected to include all previously recorded populations in this study regardless

of the last time the population was visited. In order to obtain current land owner addresses, records from respective county tax assessors’ offices were used. Permission was obtained to visit

52% of the sites on record, i.e., 32 of the 61 sites. This was in spite of multiple attempts to contact land owners. Twenty six sites were visited to assess for the presence of Georgia plume plants and to record the environmental conditions of the sites (Table 2.1). Figure 2.3 is a map

showing both the putative Georgia plume sites that were visited, as well as the sites that were not

accessed. Habitat data was collected at a total of 41 points in 28 total populations where 16

(57.1% of visited sites) were found to still contain Georgia plume plants. Twelve (42.9% of

visited sites) populations were found to contain no Georgia plume.

Active population sites that were found to still have Georgia plume plants were variable

in habitat characteristics. Habitat changed even within an individual population. Thus within the

16 active populations, twenty seven habitat recordings were taken. One third of these sites were

observed to have evidence of recent burning in the form of charred tree bark or litter. Major

habitat types included sandhill (43.8%) and flatwoods (37.5%). Other observed habitat types

included oak hammock (12.4%), ecotone between sandhill and river (12.5%), mesic (6.25%),

slope (6.25%), sandstone outcrop (6.25%), mixed bottom (6.25%), and palmetto woods (6.25%).

A mixed canopy (56%) was the prevalent type found over Georgia plume plants. Other canopy types were evergreen (25%) and deciduous (38%). Canopy closure above Georgia plume was generally open (63%) to partially closed (63%) with only two observations (13%) of a closed canopy above. Ground cover where Georgia plume was observed was mainly ground plants

(75%). Duff thickness at all populations was 0-5 cm except at one location where it was 5-30 cm.

Table 2.2 summarizes all of the conditions observed at sites confirmed to have Georgia plume plants.

Seventy five percent of the active population sites contained Georgia plume plants that

produced flowers. Numerous and prolific flower production was observed on some individuals.

This indicates that the lack of seedling recruitment observed in the wild is not due to failure of

trees to flower. Further studies are need to assess flower function, pollination, sexual compatibility, and fruit abortion as causal to lack of sexual reproduction in this species. In selected populations trees were individually assessed for reproduction. Evidence of sexual reproduction ranged from 39% to 88% of all individual Georgia plume plants.

Many of the active Georgia plume populations were composed of few individuals.

Seventy five percent of populations contained 50 or fewer individuals. Large populations composed of more than 100 individuals were very few (25% of active populations). Coates et al.

(2007) examined several different plant populations and found that as numbers of individuals within a mating population decreased the potential for inbreeding increased greatly. It has also been found that the geographic size of a population plays a role in reproduction. Bruna (2002) found that seedling recruitment was several time more likely in large continuous habitat than in smaller fragments. This is notable in later genetic analysis in that large amounts of flowers were produced but few fruit were produced indicating that populations may contain high amounts of

genetic similarity. Previous studies have also noted that fruit and seed set were higher in

populations like Big Hammock where higher numbers of individuals exist (Godt and Hamrick

1999).

Approximately 88% of active Georgia plume sites exclusively contained tree forms of the

plant. Those populations that contained small sucker forms of Georgia plume tended to be those

that had been burned. Georgia plume has been observed to sprout from root suckers (Fordham

1991). This seems to be consistent with fire damage in other ecosystem where some species

resist fire by re-sprouting after a fire event (Keeley 1987). Controlled burning as a management

tool can have a pronounced effect on tree form and reproduction. Frequency and

severity/intensity of burning need to be monitored. For example forest land in Fort Stewart is managed for military use and is frequently burned. At the several populations located within the

reservation, managers have made attempts to reduce the severity of fires around Georgia plume

plants.

Habitat characteristics were taken at sites that no longer contained Georgia plume plants to provide insight into causative factors for loss of the species at those sites. Table 2.4 gives a summary of characteristics found at inactive Georgia plume sites. Fourteen recordings were taken at inactive sites. In field visits to sites confirmed to no longer contain Georgia plume, loss of occurrence in 58% of the sites (7 of the 12) was directly attributable to human activity such as pine plantation establishment or agricultural land use. Ceballos et al. (2010) noted that agricultural land use has greatly reduced the range of native plant communities such as grasslands. Dictamnus albus is an example of another species that has suffered from considerable habitat loss due to agricultural development in Germany. Suitable habitat for the species has become rare and surrounded by agricultural fields leading to the loss of between 15 to 20 % of

populations (Hensen and Wesche 2006). Brockerhoff et al. (2008) examined native forest losses in New Zealand and found that the loss of approximately two thirds of this cover to clearing for agricultural use or other forms of development was a major contributor to loss of biodiversity.

Causes for the absence of Georgia plume at the 3 remaining sites could not be determined.

Current habitat conditions at these sites did not appear to differ appreciably from active sites.

At each active population the location of individual Georgia plume plants was recorded

using GPS. In all, 282 tree locations were recorded. At the Big Hammock, Little Hammock, and

some Fort Stewart populations locations of individuals were unable to be collected due to the

large numbers present. There are an estimated several thousand individuals present in these three

populations. These recordings will be used to calculate distances between individuals for use in

subsequent genetic analysis. This information will be used to determine whether a relationship

exists between physical distance and genetic similarity.

Aerial photographs from 2009 from ArcGIS Explorer (ESRI 2008) showed 20 unvisited

sites that were believed to no longer be active (Table 2.5). These estimates were based on clear

presence of planted pines, clear-cuts, or agricultural fields. The fact that some population

locations are very approximate may cloud these estimates as exact locations are not known. In

the same way 12 sites were estimated to possibly contain Georgia plume based on appearance of

undisturbed canopy cover (Table 2.5). Canopy cover was used as the main criterion in

determining the possible presence of Georgia plume from aerial photographs because it was

linked with presence of Georgia plume during field surveys and was clearly visible in

photographs. These results are not meant to be used instead of field observations concerning

presence of Georgia plume but were meant to prioritize such visits by avoiding locations that are

29 clearly destroyed. Aerial photographs are also a useful tool to use in conjunction with a habitat suitability model.

Habitat Suitability Model

Figure 2.4 shows a detailed area of the habitat suitability model produced from raster overlay analysis. Dark areas represent the most suitable habitat locations, those having favorable values across all habitat variables, while grey areas have favorable values in all habitat variables except one. Favorable values for the habitat variables were: 1) land cover type: deciduous forest, evergreen forest, and forested wetland; 2) mean fire return interval: 6-10 years, 11-15 years, 16-

20 years, 21-25 years, 26-30 years, 31-35 years, 36-40 years, and 41-45 years; 3) vegetation class: deciduous plantations, evergreen plantations or managed pine (can included dense successional regrowth), Atlantic Coastal Plain upland longleaf pine woodland, East Gulf Coastal

Plain interior upland longleaf pine woodland – open understory modifier, Atlantic Coastal Plain blackwater stream floodplain forest – forest modifier, and Southern Coastal Plain nonriverine basin swamp, and Southern Coastal plain hydric hammock; and 4) topographic relative moisture index: index values from 17 to 53. Areas within cities were specifically excluded from prediction because no putative populations occur within a city and anthropogenic disturbance has been observed to be detrimental to the species mainly due to habitat destruction. Bodies of water were also specifically excluded.

Out of a total area of 10786 km2, 1109 km2 (10.3% of the study area) were found to be highly favorable for Georgia plume and 2609 km2 (24.2% of the study area) were found to be moderately favorable. This corresponds to 34.5% of the study area. Unfavorable conditions were found in 7068 km2 (65.6%) of the study area. As can be seen with other threatened or endangered species, individuals are often confined to relatively small areas (Vanderpoorten,

Sotiaux et al. 2005). By building a suitability model over the entire range of Georgia plume it is possible to see areas that could be protected for the conservation of the species.

Previous work with habitat models has used input raster data with resolutions ranging from as small as 1 meter (Draper, Rossello-Graell et al. 2003) to as large as 2 kilometers

(Meinke, Knick et al. 2008). This study employed raster dataset with a resolution of 30 meters.

This is the generally accepted standard resolution for most mapping applications and is the most widely available resolution of environmental datasets. With finer resolution data better habitat suitability predictions could be made. It is also important to note that some datasets used, such as habitat type, are themselves produced from models based on sampled locations. More rigorous modeling methods are employed to develop these datasets and they are considered reliable.

Other more complex models, such as logistic regression, Classification and Regression

Tree (CART), Mahalanobis Distance, and Genetic Algorithm for Rule-set Production (GARP), have been used in previous studies with other threatened and endangered species (Pereira and

Itami 1991; Vayssieres, Plant et al. 2000; Browning, Beaupre et al. 2005; Giriraj, Irfan-Ullah et al. 2008). While these methods take the importance of different habitat characteristics into account and determine through statistical testing whether individual variables are significant, they did not seem to be applicable to Georgia plume due to extremely limited sample size. These modeling methods require presence and absence data for the study species in order to produce accurate predictions. The studies above were also conducted over much larger areas that often covered an area equivalent to several states and that captured more environmental variation.

Georgia plume is found in a comparatively small geographic area and its distribution seems to be determined by fine scale habitat variations. Kumar et al. (2009) compared several different habitat modeling techniques to predict the distribution of the diatom species Didymosphenia

geminate and found that a tool called Maxtent was best at predicting distribution. This shows that

different methods are better suited to different applications. True absence data is difficult to obtain for Georgia plume since many putative locations do not contain Georgia plume but the actual location is not known because it is based loosely on a historical description of a site.

Methods such as regression, CART, and GARP generate probability maps where each grid cell is assigned a probability that the study species is present there. Mahalanobis Distance generates a measure of how different an unknown location is when compared to a location where the study

species does exist. These methods differ from raster overlay analysis because they are not based

on binary data that result in a location being either favorable or unfavorable for the study species.

The wide variety of habitats in which Georgia plume is currently observed also makes it difficult to define any one habitat type as specifically unsuitable. Earlier descriptions of sites generally agree that Georgia plume occurs in sandy soils (Bozeman 1983) but this is very general because it is based only on one type of habitat characteristic. Simple overlay analysis, similar to the method presented, has been used previously in a study (Alexander, Craddock et al. 2005) to facilitate the location of chestnut trees. This study employed historical records of tree locations to define suitable locations based on soil series maps, topographic data, vegetation type, and aerial imagery. If it were able to be determined which habitat variables were uniquely important for the distribution of Georgia plume it would be possible to weight variables used in raster addition rather than treating them as binary (suitable vs. unsuitable).

Other studies have been able to determine distribution ranges for threatened and endangered species such as black bear, spiny pocket mice, sagebrush, and xate palms (Clark,

Dunn et al. 1993; Anderson and Martinez-Meyer 2004; Meinke, Knick et al. 2008; Penn,

Moncrieff et al. 2009). The purposes of these studies were to find land for restoration projects or

to better define ranges for prioritization of protected lands. Draper et al. (2003) used a multiple

linear regression model to determine how invasive plant species were interacting with native

species in an effort to identify areas for remediation work. These goals are generally similar to

the reasons for determining Georgia plume’s habitat which are to protect current habitat and

possibly for reintroduction planting.

2.5 CONCLUSION

This study shows that a GIS/GPS based methods can be effectively used in data

collection, record management, population inventorying etc. and conservation management of

rare and endangered plants. Georgia plume occurs in several diverse habitat conditions. Keeping

records of populations is extremely important in managing this species. This GIS based tool will enable managers to keep records in a central location and in a standard format that can be shared

with other workers or departments. These methods will also allow for records to be kept more up

to date – previous observation records dated as far back as 1979. These tools could also easily be

applied to other threatened or endangered plant species.

It was determined through field observation and aerial photography interpretation that

approximately half of previously known populations no longer exist. Nearly 75% of all

populations contain 50 or fewer individuals. These plants tend to be larger tree forms, located in

the understory, and are below open to partially open canopies.

The habitat suitability model produced revealed that approximately one quarter of the

study area is either highly or moderately favorable for Georgia plume. It shows that suitable

habitat often occurs in small, isolated patches. Based on this it appears that human activity has

isolated populations and is detrimental to the species. The model will make it possible to identify

areas be given priority for protection and to identify areas for reintroduction plantings.

REFERENCES

Alexander, M. T., J. H. Craddock, et al. (2005). "Conservation of Castanea dentata germplasm

of the Southeastern United States." Acta Horticulturae: 485-490.

Anderson, R. P. and E. Martinez-Meyer (2004). "Modeling species' geographic distributions for

preliminary conservation assessments: an implementation with the spiny pocket mice

(Heteromys) of Ecuador." Biological Conservation 116(2): 167-179.

Bozeman, J. R. (1983). Final Status Report on Elliottia racemosa Muhl. ex Ell. (Georgia plume).

G. DNR. Brunswick.

Brockerhoff, E. G., W. B. Shaw, et al. (2008). "Re-examination of recent loss of indigenous

cover in New Zealand and the relative contributions of different land uses." New Zealand

Journal of Ecology 32(1): 115-126.

Browning, D. M., S. J. Beaupre, et al. (2005). "Using partitioned Mahalanobis D-2(K) to

formulate a GIS-based model of timber rattlesnake hibernacula." Journal of Wildlife

Management 69(1): 33-44.

Bruna, E. M. (2002). "Effects of forest fragmentation on Heliconia acuminata seedling

recruitment in central Amazonia." Oecologia 132(2): 235-243.

Ceballos, G., A. Davidson, et al. (2010). "Rapid Decline of a Grassland System and Its

Ecological and Conservation Implications." Plos One 5(1).

Chafin, L. J. (2007). Field guide to the rare plants of Georgia. Athens, GA, The State Botanical

Garden of Georgia.

Clark, J. D., J. E. Dunn, et al. (1993). "A multivariate model of female Black Bear Habitat use

for a Geographic Information System." Journal of Wildlife Management 57(3): 519-526.

Coates, D. J., J. F. Sampson, et al. (2007). "Plant mating systems and assessing population

persistence in fragmented landscapes." Australian journal of botany 55(3): 239-249.

Del Tredici, P. (1987). "Lost and Found: Elliottia racemosa." Arnoldia 47(4): 3-8.

Draper, D., A. Rossello-Graell, et al. (2003). "Application of GIS in plant conservation

programmes in Portugal." Biological Conservation 113(3): 337-349.

Duncan, W. H. and M. B. Duncan (1978). "Elliottia in the Piedmont of Georgia." Castanea

43(2): 182-184.

Elliott, S. (1821). A Sketch of the Botany of South Carolina and Georgia. Charleston, SC, J.R.

Schenck.

ESRI (2008). ArcGIS Explorer. Redlands, CA.

ESRI (2008). ArcPad 7.1.1. Redlands, CA.