Site by Bennett Web & Design Co.
2011 SOUTHEASTERN NATURALIST 10(2):275–288
Magnolia grandiflora L. Range Expansion:
A Case Study in a North Carolina Piedmont Forest
Jennifer A. Gruhn1,* and Peter S. White2
Abstract - We analyzed a naturalized population of Magnolia grandiflora L. occurring
north of its native range in a temperate deciduous forest of the North Carolina piedmont.
The population was likely expanding, based on its size-class distribution; however, it had
not reached a reproductive size or age. The maximum size of established stems was 9.8
cm dbh and the maximum age, based on a subset of sampled stems, was 26 years. We
analyzed the establishment of M. grandiflora trees with respect to several environmental
variables. Climatic variables included annual minimum winter temperatures, frost-free
periods, and precipitation, and topographic variables included elevation, aspect, and
slope. We found a strong correlation between establishment and minimum winter temperatures
as well as frost-free periods, but not with other environmental variables. Magnolia
grandiflora has become naturalized north and west of its native range on the southeastern
coastal plain; among the possible causes are climate change and the increased use of
Magnolia grandiflora L. (Southern Magnolia or Bull Bay) is native only to the
coastal Southeastern United States (see Fig. 1), but has become naturalized north
and west of its native range, into New England, the Midwest, and the Southwest
(Basinger 1998, Becker 2004, USDA 2009). Though much research has addressed
the local expansion of M. grandiflora populations into fire-suppressed
pine forests within its native range (Daubenmire 1990, Delcourt and Delcourt
1977, Glitzenstein et al. 1986, Myers and White 1987), no studies have explored
the nature of its northward and westward range expansion (Becker 2004). We
analyzed a population naturalized in the piedmont region of North Carolina close
to the University of North Carolina at Chapel Hill, an area in which M. grandifl
ora has been cultivated for well over 100 years.
We were interested in this invasion in part because broad-leaved evergreens
have been used as climate indicators in Europe, due to their sensitivity to low
winter temperatures (Berger et al. 2007, Box 1981, Pott 2005, Woodward et
al. 2004). European broad-leaved evergreen species are undergoing a climateinduced
northward range expansion, potentially due to rising minimum winter
temperatures. We therefore compared the time course of M. grandiflora establishment
with local weather data for this time period.
1Department of Biology, Washington University in St. Louis, 1 Brookings Drive, Campus
Box 1137, St. Louis, MO 63130-4899. 2CB# 3280, Coker Hall, The University of North
Carolina at Chapel Hill, Chapel Hill, NC 27599-3280. *Corresponding author - JenniferAGruhn@
276 Southeastern Naturalist Vol. 10, No. 2
The pattern and process of plant invasion is an important topic of current
research in ecology and conservation biology (Fridley et al. 2007). Research has
found that invasion can be correlated with site environment, primary productivity,
disturbance, and community diversity. Of these factors, we examined whether
the establishment of M. grandiflora was correlated with topography (which is
correlated with microclimate and productivity). Although disturbance is potentially
significant, this factor is less important for a deeply shade tolerant species
like M. grandiflora. Hurricane Fran in 1996 and Hurricane Isabel in 2003 caused
moderate disturbance in our study area; however, we saw no evidence in the field
that M. grandiflora establishment was related in any way to treefall gaps, so we
did not add this factor to our study.
Our study area is located in Chapel Hill (Orange County, NC), more than 200
km northwest of the nearest native M. grandiflora population (Brunswick County,
NC). With regards to the established population, we addressed the following
questions: 1) what is the extent and size structure of the M. grandiflora population
in the study area? 2) Has the population produced reproductive individuals?
Figure 1. The shaded region is the native range of Magnolia grandiflora L. (USGS 1999).
The outlying northern locations were likely recorded from cultivation. The arrow points
to the location of the study area in Chapel Hill, Orange County, NC.
2011 J.A. Gruhn and P.S. White 277
3) Can we conclude from the answers to questions 1 and 2 that the population is
expanding? 4) Do the ages of the M. grandiflora stems correlate with a change in
yearly minimum winter temperatures, frost free periods, or annual precipitation?
and 5) Is the distribution of stems correlated with the elevation, aspect, or slope,
environmental variables which may indicate an association with microclimate or
We studied a 17.5-ha tract in the western part of 37-ha Battle Park (35.914°N,
79.04°W), which is characterized by mature upland deciduous forest dominated
by Oak and Hickory (Fig. 2). Battle Park was gifted to the University of North
Carolina in 1792 by local citizens (the University was authorized in 1789 and
opened in 1793) and became a conservation tract of the North Carolina Botanical
Garden in 2004. Though Battle Park was never farmed nor clear cut, and thus has
Figure 2. The study area, with transparent hillshade and slope relief. The circles representing
the proportional dbh of M. grandiflora stems. The shaded circles demark those
M. grandiflora cultivated outside of the study area; these trees, also represented by proportional
dbh, were fertile and potential source trees for the introduced population.
278 Southeastern Naturalist Vol. 10, No. 2
been forest since presettlement times (settlement began in 1742 in Chapel Hill),
it did experience selective harvesting for fuel and lumber, as well as understory
grazing. These practices ended by the early decades of the twentieth century.
Recreational trails have traversed the forest since the late 1800s (Battle et al.
1930) and continue to this day.
The vascular flora of Battle Park includes 642 taxa, 455 of which are native
and 187 exotic (Giencke et al. 2007). Dominant trees (from most to least
abundant) include Quercus alba L. (White Oak), Liriodendron tulipifera L. (Yellow
Poplar), Fagus grandifolia Ehrn. (American Beech), Acer rubrum L. (Red
Maple), and Carya alba (Lam. ex Poir.) Nutt. (Mockernut Hickory). Seventeen
of the 187 exotics are escapes from cultivated plants that represent species native
to the Southeast (M. grandiflora included).
The climate, as recorded by the Chapel Hill 2 W station 3.2 km away from
the study area, is characterized by a 30-year (1971–2000) average temperature
of 3.5 °C in January and 25.3 °C in July (NOAA 2005). Historical environmental
disturbances include two hurricanes (as mentioned, Hurricane Fran in 1996 and
Isabel in 2003), both of which caused moderate tree damage within the Park.
The native range of M. grandiflora stretches from the southern-most county
of North Carolina to eastern Texas (Fig. 1) (Little 1971, USGS 1999). Across this
region, populations of M. grandiflora occur in mesic to hydric areas where fires
are absent, infrequent, or of low intensity. Additionally, M. grandiflora occurs as
a late successional species on fire-suppressed sites (Delcourt and Delcourt 1977).
The species may spread by dispersal via birds, small rodents, or other mammals,
or even vegetatively through underground root shoots and trailing branches (Gardiner
M. grandiflora has a long history of cultivation in the United States and in
several countries worlwide (Missouri Botanical Garden 2011) because of its dark
evergreen leaves, large fragrant flowers, symmetrical form, and adaptability to soil
and climate in temperate as well as tropical regions. It is often cited as the most
popular cultivated broad-leaved evergreen tree (Gardiner 2000, Treseder 1978).
However, minimum winter temperatures restrict its cultivation success in the
northern United States and British Isles. Seedlings and saplings, in particular, are
susceptible to defoliation by frost damage and dehydration by winter winds when
temperatures are below -14 °C. Additionally, when subjected to a shorter reproductive
season and sooner onset of winter temperatures, older trees will not produce
fruits or flowers.
Within its native geographic range, M. grandiflora withstands minimum
winter temperatures between 6 and -12 °C. In cultivation, its has endured temperatures
as low as -18 °C. Its inability to flower when exposed to cold winter
temperatures has stimulated selection for cold hardiness by the horticulture
industry and hybridization with Magnolia virginiana L., to produce cold-hardy
genotypes (Gardiner 2000, Treseder 1978).
Within the southeastern United States, M. grandiflora cultivation dates
back to at least the 1800s. At this time, trees were planted on homesteads as
2011 J.A. Gruhn and P.S. White 279
far north as Michigan, though they did not necessarily flower (Treseder 1978).
Horticulturalists have produced over 100 cultivars of M. grandiflora and two
hybrids, the Freeman hybrids, from crosses with M. virginiana, which has a
slightly larger native range (Little 1971). Based on native trees’ resemblances
to the Freeman hybrids, there is some evidence for natural hybridization between
M. grandiflora and M. virginiana within these species’ native ranges
In Chapel Hill, M. grandiflora has been cultivated since at least 1922 (the
date of the oldest NCU herbarium specimen of the species, collected in the University’s
Coker Arboretum). Additionally, there is a reference (Battle et al. 1930)
to planted “Magnolia” trees occurring in the late 1800s on the Senlac homesite
bordering Battle Park. Without reference to species, we assume the trees to be
M. grandiflora, considering its popularity during this time period (Gardiner 2000,
Treseder 1978). Excluding the Senlac home, development of residential area
proximal to the study area began in the 1920s (Giencke et al. 2007), and planted
M. grandiflora trees are now common on these nearby properties.
We gathered data on M. grandiflora trees in the fall or winter of 2005, 2008,
and 2009. For every M. grandiflora tree over 1.4 m in height, three pieces of information
were gathered: (1) geographical position using a mapping-grade Trimble
handheld GPS unit (2- to 5-m accuracy under ideal conditions), (2) diameter
at breast height (dbh), and (3) evidence of reproduction (i.e., flowers or fruits on
the tree). We also collected this information from areas surrounding Battle Park,
including lands of the university campus and residential area (see Fig. 2) in order
to determine the size and reproductive status of nearby cultivated trees.
In order to determine the invading population’s age, we counted growth rings
for eleven trees that spanned the range of diameters we recorded in our survey.
We gathered samples from each tree by cutting three tree “cookies” (cross-sections)
at the trunk base. We then counted the number of rings for each sample,
using a dissecting microscope. Though there was minimal discrepancy between
the three cross sections per tree, the widest “cookies” generally produced rings
easiest to detect.
We analyzed local trends in climate using a comprehensive dataset of values
for daily temperatures and precipitation provided by the Southeast Regional
Climate Center of the University of North Carolina at Chapel Hill. The climate
data was recorded at the Chapel Hill 2 W weather station, NOAA Station Id:
NC311677 (35.91°N, 79.08°W). This station has an elevation of 152 m and is
located 3.2 km from the study area.
We used Microsoft Office Excel 2007 to graph the size-class distribution as
well as the estimated age of the M. grandiflora stems based on their dbh. We used
a linear trendline to generate a regression equation relating dbh to stem age; this
equation provided us an estimated year for the establishment of every stem with
a measured dbh, integral for the following climate analysis.
280 Southeastern Naturalist Vol. 10, No. 2
We also used Excel to sort and graph the climate dataset with respect to the
age of the onset of the population and establishment of stems. The graphs incorporated
the years from 1975 to 2005, to provide data both before the population
established, as well as after we have recordable data (the youngest and unmeasured
trees were most recently established). To analyze minimum temperature
thresholds of the trees, we used a pivot table to graph absolute minimum winter
temperatures for every year from 1975 to 2005. January and February were incorporated
into the previous year’s “winter” data.
For an analysis of the frost-free period of each year (commonly referred to
as the growing season), the climate dataset was reviewed for the number of days
between the last and first frosts of a given year. We estimated the temperature of
a frost using the common frost indices of 0° C (32° F) and -2.2° C (28° F).
Lastly, we averaged the daily precipitation values for all given years and also
graphed this precipitation trend relative to M. grandiflora stem establishment.
Data analysis with geographic information system
We evaluated the study area’s environmental variables using ArcGIS 9.3.1,
released by ESRI in 2008, and available through Washington University in St.
Louis. We explored M. grandiflora establishment with respect to environmental
variables. We used the USGS 2005 National Elevation Data Set entitled “twentyftdem”
for elevation data, and calculated aspect and slope from the elevation via
Spatial Analyst. We then applied the histogram tool within Geostatistical Analyst
to compare the frequency of M. gandiflora within the study area present at a given
environmental variable (stem frequency) to the frequency of the environmental
variable across the study area (pixel frequency).
Size-class distribution of stems
We recorded 438 stems of M. grandiflora in the 17.5-ha study area; the
study area was therefore characterized by 25 M. grandiflora stems/ha. Basal
area was 0.0453 m2/ha. The majority of the stems were in the smallest size
class (Fig. 3). Eighty-nine percent were less than 4 cm dbh, and 59 percent
were less than 1.5 cm dbh. The stems within the study area had a significantly
lower dbh than planted trees surrounding the study area. The maximum
dbh we observed within the study area was 9.8 cm, whereas the cultivated
trees surrounding the study area ranged from 13.8 to 33.9 cm dbh. Though
M. grandiflora occurring within the study area did not produce reproductive
structures, all trees outside of the study area bore abundant fruits and flowers.
Age of population relative to dbh
With respect to the time of M. grandiflora establishment determined by the
number of growth rings (Fig. 4), the youngest tree for which we took dbh measurements
was 5 years old and the oldest of those sampled was 26 years old.
There is a strong positive correlation between dbh and tree age (R2 = 0.84); however,
trees with low dbh appear to be more variable in age than larger stems.
2011 J.A. Gruhn and P.S. White 281
Analysis of climatic trends
Minimum winter temperatures have been rising and the growing season has
been becoming longer in Chapel Hill for the last three decades (Figs. 5, 6). There
Figure 4. Magnolia grandiflora dbh and a corresponding estimated age, as determined by
the number of growth rings of a sample of M. grandiflora trunk bases.
Figure 3. Size-class distribution of M. grandiflora stems.
282 Southeastern Naturalist Vol. 10, No. 2
Figure 5. The estimated number of M. grandiflora trees established during the initial
invasion of the population (primary axis) and the absolute minimum winter temperature
for a given year (secondary axis). The number of stems established in a given year was
based on a regression equation that calculates age as a function of dbh (see Fig. 4).
2011 J.A. Gruhn and P.S. White 283
is a trend for more trees to become established after a winter with a higher minimum
temperature and during years with longer growing-season length than for
years with colder temperatures and shorter growing seasons. Note that we could
not predict ages for the stems established from 1996 to 2005, although these years
show greater winter warmth, longer growing seasons, and a high establishment
rate for new stems.
The relationship of establishment to precipitation is less clear. While increasing
moisture and increasing establishment does occur between 1986 and 1996,
wet years between 1975 and 1986 did not have high establishment (Fig. 7).
Spatial distribution relative to elevation, aspect, and slope
The Magnolia grandiflora distribution of stems with respect to elevation,
aspect, and slope (Figs. 8, 9, 10) presented no clear trends, besides a modest underrepresentation
of M. grandiflora stems on south to southeast aspects relative
to the frequency of this aspect in the study area (Fig. 9). Counter to the hypothesis
that M. grandiflora would occur more frequently on warmer slopes, there were
Figure 6 (opposite page, bottom). The estimated number of M. grandiflora trees established
during the initial invasion of the population (primary axis) and the number of days
in a frost-free period for a given year (secondary axis).
Figure 7. The estimated number of M. grandiflora trees established during the initial
invasion of the population (primary axis) and the annual precipitation for a given year
284 Southeastern Naturalist Vol. 10, No. 2
fewer M. grandiflora on these slopes. Generally, M. grandiflora frequencies
across the range of values for the environmental variables elevation, aspect, and
slope were nearly that expected by chance, relative to the frequency of study area
pixels with the respective values.
We chose our study area in part because it occurs in an area with a long history
of M. grandiflora cultivation and therefore a long history of seed dispersal
Figure 9. Frequency histograms across a range of aspect values: the number of
M. grandiflora stems found at each aspect (above) as well as the number of 6- x 6-m
pixels (representing the entire study area) with a particular aspect value (below). Aspect
is expressed in positive degrees from 0 to 359.9, measured clockwise from north.
Figure 8. Frequency histograms across a range of elevation values: the number of
M. grandiflora stems found at each elevation (above) as well as the number of 6- x 6-m
pixels (representing the entire study area) with a particular elevation value (below).
2011 J.A. Gruhn and P.S. White 285
occurring into the area where we conducted our research. M. grandiflora within
the study area were much younger than the period of cultivation of this species on
surrounding lands and this population has not yet reached reproductive maturity.
The population profile suggests ongoing establishment and a population that will
increase in size and age, but this remains to be confirmed in future decades.
The growth rings from the tree of the largest diameter displayed an age of 26
years, and we deduce from this and the structure of the population that invasion
began in the last three decades. While M. grandiflora establishment does parallel
higher minimum winter temperatures and longer growing seasons in the climatic
record, we note that the 1930s and 1940s were also warm periods in our study
area, and it is unknown whether establishment also occurred in that time period.
There is some evidence that stems were not present in the study area during this
time period, since no herbarium specimens were collected during this period,
despite the frequent use of the area by faculty and students, the presence of much
activity through the University of North Carolina Herbarium, and several student
projects that used our study area (Giencke et al. 2007).
If, as we presume, invasion occurred in only the last three decades, then potential
causes for the recent invasion, besides the climatic warming trend, include:
(1) increased seed production of nearby planted trees, since M. grandiflora trees
become more prolific seed producers with age (Maisenhelder 1970); (2) the recent
selection or hybridization for cold-hardy M. grandiflora genotypes, which could
have been planted proximal to the study area; (3) natural hybridization of M. grandifl
ora and M. virginiana, which are both cultivated in the area; (4) further climatic
factors not considered (e.g., one explanation of the lack of invasion in the 1930s
and 1940s could be drought); and (5) changes in wind or fire disturbance. Unfortunately,
we do not have data to fully test these alternatives at present.
Figure 10. Frequency histograms across a range of slope values: the number of M. grandifl
ora stems found at each slope value (above) as well as the number of 6- x 6-m pixels
(representing the entire study area) with a particular slope value (below). Slope is expressed
in positive degrees from 0 to 90, representing degrees from horizontal.
286 Southeastern Naturalist Vol. 10, No. 2
It seems unlikely that increased seed production (alternative 1) is a cause because
of the long cultivation of Magnolia in the study area and the abundance of
fruiting specimens in herbarium collections prior to the 1970s. Use of cold-hardy
genotypes (alternative 2) is viable, as is potential hybridization (alternative 3).
Morphologically, M. grandiflora x M. virginiana hybrids are detectable by reproductive
characteristics (Gardiner 2000, Treseder 1978), which were not present
in the naturalized population. Alternative 4 would explain why invasion did not
occur in an earlier warm period and would also support that the current invasion
is due to warm conditions. Alternative 5 is unlikely because earlier windstorms
did not produce invasion, the population we observed is not confined to disturbed
patches (this is a deeply shade-tolerant species), and fires have been absent for
many decades (a change in fire frequency did not occur in the last three decades,
as fire was largely absent since at least 1900; Giencke et al. 2007).
Magnolia grandiflora cultivation has a long history near our study area, and
we do not believe that seed dispersal is limiting. However, M. grandiflora is not
yet as frequent in Duke University’s Duke Forest, a larger tract of land about 7
km distant from our study area. This tract is surrounded by much younger neighborhoods
with a shorter history of cultivation (Judson Edeburn, Duke Forest
Manager, Chapel Hill, NC, pers. comm.). Thus, it is likely that there is an interaction
between the factors causing invasion and the local history of cultivation and
therefore seed dispersal.
Future research should investigate the M. grandiflora range expansion over a
wider geographic area in order to understand the importance of cultivation history,
seed dispersal, and climatic trends. Studies should also consider whether selection
for winter hardiness and hybridization is a cause of range expansion. Lastly, future
research should determine whether range expansion will result in M. grandiflora
becoming a forest dominant, as it does in fire-suppressed forests within its native
range, as well as the impact its presence may have on other species.
Magnolia grandiflora is becoming increasingly naturalized north and west
of its native Gulf Coast range. This study has shown an association between
minimum winter temperatures and increased growing season length over the last
three decades with the increased establishment of M. grandiflora stems. The association
between climate and range expansion in a broad-leaved evergreen in a
temperate deciduous forest climate, previously investigated across broad-leaved
evergreens in Europe, may suggest that M. grandiflora will be a useful climate
indicator in the United States and other areas, given its worldwide cultivation.
Future studies should investigate whether this trend is induced by the selection
for cold-hardiness of cultivars, hybridization with M. virginiana and other
cold-hardy species, and/or the influence of environmental variables. Additional
variables, such as soil type or drought history, may provide additional correlates
with the onset of range expansion, and ideally should be tested across several
populations and wider geographic areas.
2011 J.A. Gruhn and P.S. White 287
Gratitude is expressed to Jonathan Keith Langston, Hannah Gavin, and Stephen Keith
for assistance in the field and Amanda Henley for her expertise and support with GIS. Dr.
Iván Jiménez, Trisha Consiglio, and Dr. Bob Coulter provided additional GIS materials and
assistance. The University of North Carolina Herbarium (NCU) granted access to specimens
and the Missouri Botanical Garden (MO), and Washington University in St. Louis
provided access to GIS software. Kyle Little, Bette Sumrell, and Dr. Rainer Bussmann
assisted with the charts and figures. Kevin Parentin, Ashley Glenn, Alice Kexel, Tyron
Thorson, and Lara and Russell Peterson offered insightful discussion and support.
Basinger, M.A. 1998. Notes on some naturalized woody plant species new to Illinois.
Transactions of the Illinois State Academy of Science 92:33–36.
Battle, H.B., L. Yelverton, and W.J. Battle. 1930. The Battle Book: A Genealogy of the
Battle Family in America, with Chapters Illustrating Certain Phases of its History.
Paragon Press, Montgomery, AL.
Becker. C.E. 2004. Composition, distribution, and abundance of invasive exotic plants at
Hot Springs National Park. National Park Service, Midwest Region, Heartland Network,
Republic, MO. Technical report NPS/HTLN/HOSP/P2106020870.
Berger, S., G. Söhlke, G.-R. Walther, and R. Pott. 2007. Bioclimatic limits and range
shifts of cold-hardy evergreen broad-leaved species at their northern distributional
limit in Europe. Phytocoenologia 37:523–539.
Box, E.O. 1981. Macroclimate and Plant Forms: An Introduction to Predictive Modeling
in Phytogeography: Tasks for Vegetation Science, Vol. 1. Junk Publishers, The Hague,
The Netherlands. 271 pp.
Daubenmire, R. 1990. The Magnolia grandiflora-Quercus virginiana forest in Florida.
American Midland Naturalist 123:331–347.
Delcourt, H.R., and P.A. Delcourt. 1977. Presettlement Magnolia-Beech climax of the
Gulf Coastal Plain: Quantitative evidence from the Appalachicola River Bluffs, northcentral
Florida. Ecology 58:1085–1093.
Fridley, J.D., J.J. Stachowicz, S. Naeem, D.F. Sax, E.W. Seabloom, M.D. Smith, T.J.
Stohlgren, D. Tilman, and B. Von Holle. 2007. The invasion paradox: Reconciling
pattern and process in species invasions. Ecology 88:3–17.
Gardiner, J. 2000. Magnolias: A Gardener’s Guide. Timber Press, Portland, OR.
Giencke, L., S. Keith, C.A. McCormick, A. Weakly, D. Vandermast, and P. White. 2007.
Conservation project: Battle Park. University of North Carolina, Chapel Hill, NC.
Glitzenstein, J.S., P.A. Harcombe, and D.R. Streng. 1986. Disturbance, succession, and
maintenance of species diversity in an east Texas forest. Ecological Monographs
Little, E.L., 1971. Atlas of United States Trees, Volume 1. Conifers and Important Hardwoods.
USDA Forest Service, Misc. Publ. No. 1146. Washington, DC.
Maisenhelder, L.C. 1970. Magnolia. USDA Forest Service, American Woods FS-245.
Washington, DC. 7 pp.
Missouri Botanical Garden. 2011. Tropicos.org. Available online at http://www.tropicos.
org. Accessed 17 January 2011.
Myers, R.L., and D.L. White. 1987. Landscape history and changes in sandhill vegetation
in north-central and south-central Florida. Bulletin of the Torrey Botanical Club
288 Southeastern Naturalist Vol. 10, No. 2
National Oceanic and Atmospheric Administration (NOAA). 2005. Annual summary for
North Carolina. Available online at http://www.noaa.gov. Accessed 26 October 2009.
Pott, R. 2005. Allgemeine Geobotanik. Springer, Berlin, Germany. 652 pp.
Treseder, N.G. 1978. Magnolias. Faber and Faber, Boston, MA.
US Department of Agriculture, Forest Service. 1974. Seeds of Woody Plants in the
United States. C.S. Schopmeyer (Tech. Coord.). US Department of Agriculture, Agriculture
Handbook 450. Washington, DC. 883 pp.
US Department of Agriculture, NRCS (USDA). 2009. The PLANTS Database. Available
online at http://plants.usda.gov. Accessed 26 May 2009. National Plant Data Center,
Baton Rouge, LA.
US Geological Survey (USGS). 1999. Digital representation of “Atlas of United States
Trees” by Elbert L. Little, Jr. US Geological Survey Professional Paper 1650. Denver
Federal Center, Denver, CO.
Woodward, F.I., M.R. Lomas, and C.K. Kelly. 2004. Global climate and the distribution
of plant biomes. Philosophical Transactions of the Royal Society: Biological Sciences