The Continuing Population Decline of Torreya taxifolia Arn.

Abstract

Torreya taxifolia is a federally endangered dioecious conifer endemic to ravine bluffs along the east side of the Apalachicola River in Gadsden and Liberty Counties, Florida and Decatur County, Georgia (plus one outlying population in Jackson County, Florida). Torreya taxifolia experienced a catastrophic population decline, presumably fungal in origin, in the late 1950's. Within 10 years virtually no adults remained in the wild, a situation that persists to the present. Size measurements of trees indicate that the mean length of the longest stem is less than one meter. In a census population of over 100 trees, 10% died in four years, mostly from small size classes. Extension of terminal buds along the primary stem was infrequent, with only 47% of all trees experiencing any growth in length during the four year census period. With 32% of all trees losing their primary stem during the census interval, the mean size of individuals within the population is decreasing. Most individuals are multi-stemmed and continue to carry symptoms of disease. Although no causative agent for the decline has been identified, the most prevalent symptoms of disease are needle spots, needle necrosis and stem cankers. Growth fate does not appear to be strongly related to habitat or the incidence of disease symptoms, although stem mortality was higher among trees with heavy foliar pathogen loads. With no sexually mature trees and no seed production, this species faces continued decline.

Full text

Estimating the magnitude of decline of the Florida torreya
(Torreya taxifolia Arn.)
Mark W. Schwartz
a,
*, Sharon M. Hermann
b
, Philip J. van Mantgem
a
a
Environmental Science and Policy, One Shields Avenue, University of California, Davis, CA 95616, USA
b
Tall Timbers Research Station, 13093 Henry Beadel Drive, Tallahassee, FL 32312, USA
Received 23 June 1999; received in revised form 16 August 1999; accepted 20 November 1999
Abstract
Torreya taxifolia is a coniferous tree that is endemic to the 35 km stretch of blus and ravines along the east side of the Apa-
lachicola River in northern Florida and adjacent southern Georgia. This formerly locally abundant tree declined during the 1950s
and 1960s as a result of disease and is currently on the US Endangered Species list. For sparsely distributed species it can often be
dicult to determine both current and historic population sizes. Historical descriptions of the distribution (203 km
2
) and relative
abundance (14.2% of dominant ravine trees) of T. taxifolia are used along with current measures of forest structure to estimate the
pre-decline population density (30 trees/ha) and size (0.3±0.65 million individuals). Survey information from ®ve extant stands is
used to estimate current population size (500±4000 individuals). The surveys were conducted in areas with known high tree den-
sities such that a simple extrapolation to the entire distribution would produce a gross over-estimate of population size. We there-
fore use a variety of assumptions to produce a range of estimates for total population sizes. Regardless of the particular model, our
estimates suggest that T. taxifolia has lost at least 98.5% of its total population size since the early 1900s. We discuss these results in
relation to the potential diculties likely of restoring sustainable populations of this species. #2000 Elsevier Science Ltd. All rights
reserved.
Keywords: Torreya taxifolia; Endangered species; Population decline; Conifer; Florida
1. Introduction
Torreya taxifolia Arn. (Florida torreya) is in danger
of extinction due to a catastrophic decline that began in
the late 1950s (Godfrey and Kurz, 1962). Common
within the ravine forests along the Apalachicola River
during the 1940s and 1950s, Godfrey and Kurz (1962)
reported that the species was on the brink of extinction.
Since 1962 there have been no observed canopy sized
trees, nor mature seed producing trees (USFWS, 1986;
Schwartz and Hermann, 1993a). Perhaps due to this
dramatic population crash, T. taxifolia was one of the
®rst plants to receive federal protection under the
Endangered Species Act (USFWS, 1984). Despite a
small population size, continued population losses, and
no known seed production for over 30 years, extinction
of T. taxifolia has not yet occurred (Schwartz and Her-
mann, 1993a; Schwartz et al., 2000).
The catastrophic decline of this species was remark-
able for two reasons. First, this species was formerly
common (Chapman, 1885; Harper, 1914; Reinsmith,
1934) within what was one of the narrowest distribu-
tions of any tree in North America that reached canopy
stature (Kurz, 1938; Little, 1971). Second, habitat loss
was not a factor in the decline: the vast majority of the
original ravine forest habitat remains. In fact, T. taxi-
folia is one of relatively few endangered plants recog-
nized to be primarily threatened by disease (Wilcove et
al., 1998). Attempts to isolate a primary disease agent
have identi®ed many associated fungi, some of which
may be pathogenic (Al®eri et al., 1994). This work,
however, has failed to isolate a clear cause of the decline
(Schwartz and Hermann, 1993b).
Population projections predict steadily declining
numbers, but that T. taxifolia is not likely to go extinct
within the next 50 years (Schwartz et al., 2000), sug-
gesting that there is still time to restore the species in the
wild. It is dicult, however, to set appropriate goals for
the recovery process without some historical target for
0006-3207/00/$ - see front matter #2000 Elsevier Science Ltd. All rights reserved.
PII: S0006-3207(00)00008-2
Biological Conservation 95 (2000) 77±84
www.elsevier.com/locate/biocon
* Corresponding author.
E-mail address: mwschwartz@ucdavis.edu (M.W. Schwartz).

population size. Historical data is frequently used for
setting targets in ecological restoration of community
structure and ecosystem processes (e.g. Vankat and
Major, 1977; Brown and Swetnam, 1994; Minnich et al.,
1995, Fule
Âet al., 1997). This historical information has
been less frequently used to help de®ne recovery goals
for endangered species. Our goal for this research was
to use our ®eld survey data along with available forest
and habitat data in order to (1) estimate current total
population size; (2) focus attention on information gaps
in making a better estimate of current population size;
(3) estimate the magnitude of the decline; and (4) set
targets for restoration and recovery of T. taxifolia.
1.1. Species background
The Florida torreya is a dioecious coniferous tree
endemic to blus and ravines along the east side of a 35
km stretch of the Apalachicola River in the central
panhandle of Florida (Fig. 1). After its discovery in the
mid-1800s, T. taxifolia was used for fenceposts, shin-
gles, and as a fuel for riverboats on the Apalachicola
(Chapman, 1885; Gray, 1889). Many of the largest trees
were harvested during the ®rst half of this century (A.
Gholson, pers. comm.).
While T. taxifolia once reached 15±20 m in height at
maturity (Godfrey, 1988), today most naturally occur-
ring individuals consist of 1±3 immature stems less than 2
m tall (Schwartz and Hermann, 1993a). Although disease
was thought to have caused the decline, low light condi-
tions in much of the existing habitat may, at present,
serve as a principal proximate factor limiting recovery
(Schwartz et al., 1995, Schwartz and Hermann, 2000).
2. Methods
2.1. Range size estimate
Historic accounts of the distribution of T. taxifolia
describe the species as inhabiting the ravine slopes along
the eastern side of the Apalachicola River from Bristol,
FL to just across the Florida±Georgia state line, north
of Chattahoochee, FL (Chapman, 1885; Kurz, 1938;
Savage, 1983; Godfrey, 1988). We approximated this
distribution as encompassing the entire watershed of all
creeks that enter the Apalachicola River from the east
between Bristol, FL (Liberty County) and the one drai-
nage north of Chattahoochee, FL in southern Georgia
(Fig. 1). The northernmost wild individual currently
grows in Georgia within 200 m of the Florida state line.
The historically small outlying population west of the
Apalachicola, consisting of ®ve trees in 1989 (M. Schwartz
and R. Nicholson, pers. obs.), was not included in this
analysis. Kurz (1938) reported this outlying population
as being restricted to one stand with few trees.
A geographical information system (GIS; Anon.,
1989) was used to digitize 1950s USDA aerial photo-
graphs with the following habitat classi®cation: (1)
towns; (2) agricultural land; (3) upland pine forest; (4)
ravine forest; and (5) bottomland forest. Aerial photos
from the 1950s were used because upland logging at that
time maximized our ability to distinguish upland from
ravine habitats. There has been very little ravine habitat
loss since that time. According to historical records as
well as our own ®eld surveys this species is restricted to
ravine forests (Kurz, 1938; Schwartz, 1990). Thus, we
used the area classi®ed as ravine forests as an estimate
of potential habitat for T. taxifolia.
2.2. Current population size estimate
The current population size of T. taxifolia was di-
cult to estimate because the species is sparsely and
patchily distributed. We based our estimates of total
Fig. 1. Estimated distribution of Torrega Taxifolia including an out-
lying population in the vicinity of Lake Ocheessee. Habitat features
were digitized from 1953 and 1955 USDA aerial photographs using
ARC/INFO (ver. 5.0). Within this distribution T. taxifolia is restricted
to ¯oodplain and ravine habitats. Habitats sampled for current popu-
lation estimates are noted: (A) Lake Seminole; (B) Flat Creek; (C)
Torreya State Park; (D) Apalachicola Blus and Ravines Preserve
(north section); and (E) Apalachicola Blus and Ravines Preserve
(south section).
78 M.W. Schwartz et al. / Biological Conservation 95 (2000) 77±84

population size on thorough surveys of ®ve populations
(Fig. 1). Teams of between two and 12 people scoured
each survey area and ¯agged all individuals encoun-
tered. Most of these sites have been re-visited each year
for up to 10 years for demographic monitoring and few
(<20) additional individuals have been found that were
missed during the initial surveys. We estimated local
density using area estimates for surveyed regions. We
estimated total current population size using a range of
assumptions regarding the degree to which the surveyed
regions re¯ect the entire population. In our most liberal
estimate, we used our estimated density of surveyed sites
as the density throughout the range. We know, how-
ever, that this estimate is inordinately large. The sur-
veyed sites were chosen for demographic studies
speci®cally because we knew that they contained high
densities of T. taxifolia. The surrounding habitats, in
general, contain much lower densities of trees.
At the other extreme, our most conservative estimate
assumed that our information regarding the locations of
populations is thorough such that our surveys tallied
most individuals. Likewise, we know this to be overly
conservative because we have, on occasion, observed of
trees outside of the survey areas (Schwartz and Her-
mann, pers. obs.). Con®dent that the true population
lies somewhere in between these extremes, we used
arbitrary scalars (1.0, 0.5, 0.1, 0) to estimate densities of
individuals in areas where we have not surveyed. Fur-
ther, we dierentiated areas not surveyed as either close
to (<3.5 km) or far (>3.5 km) from the Apalachicola
River. We made these distinctions based on two obser-
vations. First, we have observed a few scattered trees
elsewhere within approximately 3.5 km of the Apa-
lachicola River. Second, we have never encountered a
tree further than 3.5 km from the banks of the river
during our ®eld surveys (Schwartz and Hermann, pers.
obs.). There are, however, several signi®cant tracts of
private land for which there is no survey data. We then
employed a variety of combinations of scalars for
unsurveyed habitats to obtain a range of current popu-
lation size estimates.
2.3. Pre-decline population size estimate
Harper (1914) estimated pre-decline forest composi-
tion of the region. Traveling by horseback, foot or train
across northern Florida, Harper tabulated the fre-
quency of trees in meandering transects. Harper then
aggregated transects within biogeographical regions in
northern Florida in order to estimate relative fre-
quencies within regions. Harper's (1914) data for the
Apalachicola River Blus and Ravines region, although
not ideal for calculating historic ranges and population
densities, is the only quantitative pre-decline forest
inventory in this area. In order to convert Harper's
(1914) relative frequencies to absolute densities we used
assessments of current ravine forest structure (Schwartz,
1990; Kwit et al., 1998) to estimate tree densities within
this habitat.
Similar to our estimation of current population size,
we calculated a range of estimated pre-settlement
population size based on diering assumptions regard-
ing how pre-settlement population densities varied
across its range. Early descriptions of the distribution
are incomplete regarding the extent to which T. taxifolia
was found throughout ravine forests of the Apalachi-
cola River (Chapman, 1885; Harper, 1914; Kurz, 1938).
We do not know the extent to which density may have
varied across the range. Based on current populations
we might speculate that the species may also have for-
merly been disproportionately abundant in ravines
relatively close to the Apalachicola River. In our upper
end estimates we assumed that population densities
were uniformly high. In our lower estimates we assumed
that historic populations, like the current population,
decreased in density with increasing distance from the
Apalachicola River. We varied these estimates with the
same scalars used for our estimates of current popula-
tion size. For all estimates we assumed that current
forest structure was an adequate surrogate for historic
tree densities within dierent size classes.
We used a variety of methods to estimate ratios of
young trees to mature trees in order to deal with the
problem that arises from Harper's claim that he restric-
ted his measured frequencies to canopy trees. In our
simplest estimate we multiplied the relative frequency of
T. taxifolia, as assessed by Harper, to modern estimates
of tree densities from Kwit et al. (1998) in order to esti-
mate the population size of trees >2 cm diameter at
breast height (dbh). Kwit et al. (1998) surveyed over 4
ha of ravine forest using observations of all trees >2 cm
dbh in 6 m wide belt transects (n=75) spanning from
ravine top to bottom.
A more realistic pre-decline population estimate
required that we accommodated Harper's claim that he
considered only canopy trees. Being unclear on what
Harper may have meant by ``mature'', we estimated
population size of trees greater than 25 cm and then cal-
culated the density of these trees using data from Kwit
et al. (1998). This required multiplying Harper's relative
densities to Kwit et al.'s (1998) absolute density of large
trees. We then estimated the ratio of the density of trees
from 2 to 25 cm dbh to those greater than 25 cm dbh
using modern forest structure data (Schwartz, 1990).
Summing these values we estimated population sizes for
all trees >2 cm dbh. We assessed the degree to which
population estimates varied with diering de®nitions of
canopy tree size classes by replicating this estimate using
a 15 cm de®nition of a canopy tree.
The lack of extant T. taxifolia greater than 10 cm dbh
limits the accuracy of our historical population esti-
mates that include small trees. Ravine tree species vary
M.W. Schwartz et al. / Biological Conservation 95 (2000) 77±84 79

with respect to size class distributions (Schwartz, 1990).
We do not, however, have any method with which to
predict historic size class distributions of T. taxifolia.
Lacking size structure data for T. taxifolia we assumed
that the current size structure (the ratio of small to large
trees) for all trees can be used to estimate the abundance
of small T. taxifolia given an estimate of the density of
larger individuals.
Our estimates do not account for potential dierences
in forest structure between historic and modern forests
as a result of logging. We assumed that the current for-
est has comparable densities of trees greater than 25 and
15 cm dbh as early in this century. This assumption may
result in an overestimation of historical canopy tree
densities because selective logging within ravine forests
during the 1950s likely reduced the density of very large
trees (e.g. >50 cm dbh) and increased the density of
smaller trees (Schwartz, 1990). The magnitude of this
eect is unknown, but likely to be small. Although some
of our models may over-estimate total historic popula-
tion size, these estimates remain conservative given the
fact that we did not include the likely larger population
of individuals <2 cm dbh.
These methods resulted in a range of both modern
and pre-settlement population estimates by varying our
assumptions. Asserting that the actual historic and
modern population sizes fall within the range of our
estimates, we used these values to examine the magni-
tude of the decline in T. taxifolia.
3. Results
3.1. Range size
Our GIS coverage of the range of T. taxifolia is
55,239 ha (Fig. 1). Of this region, ravines occupy 36.8%
of the total area. This results in an estimate of 20,370 ha
of ravine habitat potentially available for T. taxifolia
(Fig. 2). There has been virtually no development of
ravine habitats, and thus the current available habitat is
essentially the same as that in the 1950s.
Historic distribution descriptions are vague regarding
the speci®c drainage that de®nes the northern distribu-
tion limit (Savage, 1983). Our 20,370 ha estimate of the
range size is generous with respect to including all of the
possible ravine habitats. Among these ravine habitats, a
total of 9,650 ha are within 3.5 km of the ¯oodplain and
considered ``near''. The remaining 10,664 ha is con-
sidered far from the river for the purposes of de®ning T.
taxifolia population size.
3.2. Current population size
Current population size was estimated in several
ways. Since 1989 we have documented the existence of
365 trees through surveys of ravine forest habitat. These
trees include those surveyed by volunteers for Torreya
State Park and The Nature Conservancy as well as by
the authors and their research assistants. Surveyed trees
were observed in ®ve separate populations (Fig. 1) with
a total area of approximately 1825 ha (0.2 trees/ha;
Table 1). An assumption of uniform density throughout
the range results in an estimated population size of 4063
trees (20,314 ha0.2 trees/ha), representing the highest
possible estimate for current population size. A handful
of scattered individuals have been encountered outside
surveyed regions by the authors and others. Combining
the survey population (365 trees) along with an
assumption of low density (10% of survey density) in
unsurveyed regions near the river (7825 ha0.02 trees/
ha) and no trees far from the river (10,664 ha0.0 trees/
ha) results in our lowest estimate of current population
size at 521 trees.
Our population surveys focused on tracts known to
have large populations. Owing to a lack of survey data
Fig. 2. Estimated area of ®ve habitat classes in the Apalachicola Blus
and Ravines biogeographic province. Ravines represent the only
available habitat for Torreya taxifolia and are split into that portion of
ravines within 3.5 km of the river, where trees have been located. No
known trees exist outside this 3.5 km region of ravines.
Table 1
Current population density estimates for Torreya taxifolia based on
thorough surveys of ®ve local populations
Location Trees Approximate
area (ha)
Trees/ha
1. Apalachicola Blus and
Ravines Preserve, South
98 675 0.145
2. Apalachicola blus and
Ravines Preserve, North
87 500 0.174
3. Torreya State Park 125 300 0.417
4. Flat Creek 30 200 0.150
5. Woodru dam 25 150 0.167
Total 365 1825 0.200
80 M.W. Schwartz et al. / Biological Conservation 95 (2000) 77±84

from random locations we are forced to estimate popu-
lation size with a series of ad hoc rules based on perso-
nal experience. For example, if we assume the observed
density of 0.2 trees per ha over the half of the ravine
habitat nearest the river and no trees in the 10,664 ha
furthest from the Apalachicola River (a region from
which we know of no extant trees), the population esti-
mate is 1930 trees.
If we assume that: (a) we have surveyed most of the
good habitats such that; (b) densities in the unsurveyed
regions near the river (7825 ha) average half that of
surveyed regions; and (c) there are few trees (e.g. 0.1
times surveyed density) in the 10,664 ha furthest from
the Apalachicola River; then our population estimate is
1361 extant trees. Other combinations of reduced den-
sity estimates in unsurveyed regions near and far from
the river resulted in a variety of estimates of current
population size (Table 2).
Any of the estimates between 500 and 4000 indivi-
duals may be closest to the true population size. None-
theless, our experience of sparse trees outside survey
areas and no known trees far from the river leads us to
place the greatest con®dence in the estimate of 1361
extant trees (i.e. 0.5 times surveyed density in unsur-
veyed regions near the river and 0.1 times surveyed
density far from the river). We recommend treating this
as an upper limit to the expected extant population size.
Without further extensive ®eld surveys, we can not test
the assumptions of the various estimates.
3.3. Pre-decline population size
Dierent interpretations of the historical data lead to
dierent pre-settlement population estimates. We pre-
sent the results for four estimates. First, we assumed
that Harper's (1914) estimates of relative abundance
re¯ect all sizes of trees. Harper estimated a relative of
density of 4.0% for T. taxifolia, out of a cumulative
density of 26.7% total for the ®ve dominant ravine for-
est trees (T. taxifolia, plus Magnolia grandi¯ora,Pinus
glabra,Fagus grandifolia, and Ilex opaca; Table 3). The
remainder of trees that comprise the bulk of Harper's
estimates are either upland ¯atwood or wetland species
and were not counted in the ravine forest component.
Thus, Harper estimated that T. taxifolia represented
14.98% (4.1/26.7) of these ®ve species occurrences. Kwit
et al. (1998) observed a density of 212.6 trees/ha of these
®ve predominant ravine species (Table 3). We might
then assume that 14.98% of these 212.6 trees/ha, or
30.85 trees/ha, were T. taxifolia. This results in a pre-
decline population estimate of 646,950 trees >2 cm dbh
(20,314 ha30.85 trees/ha).
We next used Harper's (1914) assertion that his rela-
tive densities were based on canopy trees. Schwartz
(1990) observed a total of 30.17 trees/ha of the ®ve
dominant ravine species in size classes greater than 25
cm dbh in four separate ravine habitats (Table 3). This
results in an estimate for T. taxifolia of 4.52 trees/ha
(14.98% of 30.17 trees/ha) and a range-wide population
estimate of 91,808 large trees. Assuming that the pre-
settlement distribution of small to large trees is similar
to current forest structure, then we expect approxi-
mately 7.1 times as many trees between 2 and 25 cm dbh
as those >25 cm dbh (Table 3) and an estimated popu-
lation of 651,840 trees. We repeated this estimate using
a cut-o of 15 cm dbh, yielding an estimate of 147,739
canopy trees and a total population of 650,054 (Table
2).
We then used the approximate mean of these three
population estimates (650,000 trees) and applied a vari-
ety of scalars (0.5, 0.1, 0) to the estimate for populations
far from the Apalachicola River. The resulting popula-
tion estimates ranged from 325,000 to 487,500 (Table 2).
In summary, the pre-decline population of T. taxifolia
was likely to be between 325,00 and 650,000 individuals
>2 cm dbh. As before, we suspect that historic popu-
lation densities were low in ravine habitats far from the
river such that 357,500 individual trees is our preferred
estimate. These population estimates do not include
seedlings, which would likely increase population sizes
substantially.
Table 2
Predictions of current and historic population size of Torreya taxifolia
Density assumption (scalar) Population size estimate
Near river
a
Far from river
a
Current Historic
1. High (1.0) High (1.0) 4063 650,000
a. assuming Harper counted
all trees
646,950
b. assuming Harper
counted trees >25 cm dbh
651,840
c. assuming Harper
counted trees >15 cm dbh
650,054
2. High (1.0) Moderate (0.5) 2996 479,400
a. moderate (0.5) in
unsurveyed near regions
b
2214
3. High (1.0) Low (0.1) 2143 342,900
c
a. moderate (0.5) in
unsurveyed near regions
b
1361
c
b. low (0.1) in unsurveyed
near regions
b
735
4. High (1.0) None (0.0) 1930 308,800
a. moderate (0.5) in
unsurveyed near regions
b
1148
b. low (0.1) in unsurveyed
near regions
b
521
a
The 9650 ha of ravine habitat within 3.5 km of the river ¯ood-
plain were classi®ed as near river while the remaining 10,664 ha of
ravines were classi®ed as far from the river.
b
Field surveys encompassed 1825 ha of habitats near the Apa-
lachicola River, leaving 7825 ha unsurveyed near the river.
c
Predicted to be the most accurate estimate. See text for details and
rationale.
M.W. Schwartz et al. / Biological Conservation 95 (2000) 77±84 81

3.4. The magnitude of decline
Quite obviously, these are rough approximations of
current and past population sizes. Combining the smal-
lest estimate of the pre-settlement population (300,000
trees) with the largest estimate of the current population
(4000 trees) we estimate that the population has
declined by a minimum of 98.7% during this century
(Fig. 3). Using our preferred estimates of 340,000 pre-
decline trees and a current population of 1350 trees,
the decline resulted in the loss of 99.6% of the popula-
tion. Virtually all extant trees fall below the 2 cm mini-
mum size cut-o for historic population size estimates.
If we had the ability to include trees <2 cm dbh in his-
toric population estimates, then the estimated magni-
tude of decline would be well over 99.9% of the pre-
settlement population.
4. Discussion
There was a wide range encompassed by our estimates
of both current and historic population sizes. Despite
this variation in estimates it is clear that the magnitude
of decline from historic to current population levels was
enormous (99%). This conclusion holds regardless of
the choices of pre-decline and present-day population
estimates. Unfortunately, we cannot place con®dence
intervals around these estimates.
Describing the geographic distribution and assessing
the pre- and post-decline population sizes of T. taxifolia
required several assumptions. The GIS information
used to describe the range of potential habitats allowed
us to make explicit our assumptions concerning the dis-
tribution of abundance of T. taxifolia. For example, we
assumed that the 1950s aerial photos accurately descri-
bed both current and historic distributions of available
habitat. Our speci®cation of the potential distribution
within watersheds suggests locations for additional sys-
tematic population surveys to be conducted in order to
re®ne our population estimates. Our scalars represent a
hypothesis for potential variation in observed density.
The extent of the decline leads to several important
considerations for the management and recovery of the
species. First, we expect that a population decline of this
magnitude was associated with a signi®cant loss of
genetic variation. T. taxifolia is characterized by low
genetic variability, although this low variability appears
to pre-date the current decline (Schwartz, 1993). A spe-
cies with little existing variation would likely incur smal-
ler losses in variability through a dramatic population
decline. Nonetheless, this population decline represents
a potential for a genetic bottleneck that may create
additional complications for recovery eorts. It remains
to be seen if genetic issues hamper recovery eorts for
the species. There have been no observed seed set pro-
blems among those few trees growing in captivity that
produce seed (R. Determan, pers. comm.).
A second management implication is that negative
density dependence (Allee eects) might limit population
recovery of T. taxifolia given its extremely low popula-
tion density compounded by its dioecious life history.
Seed production in the wild will require a minimum of
Table 3
Pre- and post-decline population estimates of dominant ravine tree species
a
Species Harper's (1914) relative frequency Densities (#/ha) from ravine transects
All >15 cm dbh >25 cm dbh
Magnolia grandi¯ora 9.5 41.9 21.16 16.66
Pinus glabra 5.6 25.7 5.85 4.96
Fagus grandi¯ora 4.1 44.7 12.60 7.65
Torreya taxifolia 4.0 0.00 0.00 0.00
Ilex opaca
b
3.5 100.3 9.00 0.90
Sum 26.7 212.6 48.61 30.17
a
Pre-decline estimates are expressed as a modi®ed relative frequency (Harper, 1914). Post-decline forest estimates are absolute densities and
derived from ravine transect data described in Kwit et al. (1998) and Schwartz (1990).
b
Typically small understory trees.
Fig. 3. Estimates of historic and current population sizes for Torreya
taxifolia.
82 M.W. Schwartz et al. / Biological Conservation 95 (2000) 77±84

one mature male and one mature female within close
proximity to one another. Given the low density of
individuals even in high quality stands, where current
density is highest, population augmentation is likely to
be required in order to achieve sustainable densities. It
will be dicult to assess, however, whether habitats
currently housing low densities of trees historically con-
tained few trees or whether the low density at present is
a result of habitat selectivity in the recent decline.
We make three recommendations for eorts at popu-
lation augmentation. First, planting should focus on
regions where population density is currently observed
to be high (i.e. near the Apalachicola River). Second,
speci®c planting locations should be isolated from
extant populations to avoid accidental contamination
of either set of trees with pathogens. Given that current
populations are patchily distributed, it is possible to use
ravine systems that currently contain trees, yet plant
new trees more than 100 m from existing individuals.
Third, historical population size estimates provide the
best target density for management at this point in
time.
The third implication relates more generally to
restoration of endangered plant species. Unlike many
federally listed endangered species the habitat of T.
taxifolia is intact and not threatened at present. The
possibility of full recovery exists. However, restoring
population levels to even the most conservative estimate
of pre-decline population size would require a long-term
level of commitment unlikely to be seen for this species
or any other non-vertebrate. For example, a 1995 report
of total expenditures on endangered species lists funding
of $1500 on T. taxifolia, placing it 884th out of 926
species in funding (USFWS, 1998). There were no
plants in the 100 best funded endangered species
(USFWS, 1998). Nonetheless, a serious recovery eort
will require a signi®cant increase in ®nancial support.
Our research, at least, provides an initial target popula-
tion size for recovery.
This discussion of recovery presumes that the current
decline can be reversed. The feasibility of recovery for
T. taxifolia in its natural habitat hinges on our ability to
learn whether it is possible for the current population to
reproduce by seeds. Management actions cited by the
recovery plan focus primarily on better understanding
of the role of disease on survivorship and recruitment
(USFWS, 1986). Although much work has been done
on the pathogens of the species (e.g. Al®eri et al., 1994),
a primary pathogen of the decline has not been identi-
®ed (Schwartz, 1993).
Initial planting of trees from cuttings suggest that
population augmentation is possible (Schwartz and
Hermann, unpubl. data). Despite this potential opti-
mism, we have yet to observe any trees reaching repro-
ductive maturity in the wild. It is critical to determine if
changes in the ravine habitats along the Apalachicola
River have reduced the potential for recovery of T.
taxifolia. Owing to the magnitude of the decline, how-
ever, full recovery of T. taxifolia will be slow under even
the most optimistic scenario.
Acknowledgements
We thank the numerous volunteers that have helped
to survey populations of T. taxifolia over the past dec-
ade (Boy Scouts at Torreya State Park, Nature Con-
servancy Volunteers at Apalachicola Blus) as well as
W. Baker, B. Fay, and A. Gohlson who were instru-
mental in locating populations. We also thank G.
Wright and two anonymous reviewers for helpful com-
ments on earlier drafts of this manuscript.
References
Al®eri Jr, S.A., Langdon, K.R., Kimbrough, J.W., El-Gholl, N.E.,
Wehlburg, C., 1994. Diseases and disorders of plants in Florida.
Bulletin #14. Florida Dept. of Agriculture and Consumer Services,
Gainesville, FL.
Anon., 1989. ARC-INFO ver. 5.0. Environmental Systems Research
Institute, Redlands, CA, USA.
Brown, P.M., Swetnam, T.W., 1994. A cross-dated ®re history from
coast redwood near Redwood National Park, California. Canadian
Journal of Forest Research 24, 21±31.
Chapman, A.W., 1885. Torreya taxifolia Arn., a reminiscence. Bota-
nical Gazette 10, 251±254.
Fule
Â, P.Z., Covington, W., Moore, M.M., 1997. Determining refer-
ence conditions for ecosystem management of Southwestern pon-
derosa pine forests. Ecological Applications 7, 895±908.
Godfrey, R.K., 1988. Trees, Shrubs, and Woody Vines of Northern
Florida and Adjacent Georgia and Alabama. Univ. of Georgia
Press, Athens.
Godfrey, R.K., Kurz, H., 1962. The Florida torreya destined for
extinction. Science 136, 900±902.
Gray, A. 1889. A pilgrimage to Torreya. In: Sargent C.S. (Ed.), Sci-
enti®c Papers of Asa Gray, Vol. 2. Houghton Miin Co., New
York.
Harper, R.M., 1914. Geography and Vegetation of Northern Florida
(Sixth Annual Report). Florida Geological Survey.
Kurz, H., 1938. Torreya west of the Apalachicola River. Proc. Fla.
Acad. Sci. 3, 66±77.
Kwit, C., Schwartz, M.W., Platt, W.J., Geaghan, J.P., 1998. The dis-
tribution of tree species in steepheads of the Apalachicola River
Blus, Florida. Bull. Torrey Bot. Club 125, 309±318.
Little Jr, E.L., 1971. Atlas of United States Trees. Vol. 1: Conifers and
Important Hardwoods. Miscellaneous Publication no. 1146. Dept.
of Agriculture, Washington, DC.
Minnch, R.A., Barbour, M.G., Burk, J.H., Fernau, R.F., 1995. Sixty
years of change in Californian conifer forests of the San Bernardino
Mountains. Conservation Biology 9, 902±914.
Reinsmith, W.H., 1934. Exploring for torreya trees in the Apalachi-
cola blu country. Unpublished report to Florida Forest Service,
Tallahassee.
Savage, T., 1983. A Georgia station for Torreya taxifolia Arn. sur-
vives. Florida Scientist 46, 62±64.
Schwartz, M.W., 1990. Conserving forest diversity in northern
Florida: from landscapes to populations. PhD dissertation, Florida
State University, Tallahassee.
Schwartz, M.W., 1993. Allozyme variation of the endangered Florida
M.W. Schwartz et al. / Biological Conservation 95 (2000) 77±84 83

torreya (Torreya taxifolia). Canadian Journal of Forest Research
23, 2598±2602.
Schwartz, M.W., Hermann, S.M., 1993a. The continuing population
decline of Torreya taxifolia Arn. Bulletin of the Torrey Botanical
Club 120 (3), 275±286.
Schwartz, M.W., Hermann, S.M., 1993b. The population ecology of
Torreya taxifolia: habitat evaluation, ®re ecology, and genetic
variability. Final report, The Florida Game and Fresh Water Fish
Commission, Nongame Wildlife Program, Tallahassee, FL.
Schwartz, M.W., Hermann, S.M., Vogel, C., 1995. The catastrophic
loss of Torreya taxifolia: assessing environmental induction of dis-
ease hypotheses. Ecological Applications 5, 501±516.
Schwartz, M.W., Hermann, S.M., van Mantgem, P.J. in press. Popu-
lation persistence in Florida torreya (Torreya taxifolia Arn.) Ð
slowly declining toward extinction. Conservation Biology.
USFWS (United States Fish and Wildlife Service) 1984. Endangered
and threatened wildlife and plants: ®nal rule to determine Torreya
taxifolia (Florida torreya) to be an endangered species. Federal
Register 49(2783-2786).
USFWS (United States Fish and Wildlife Service) 1986. Florida tor-
reya Torreya taxifolia recovery plan. U.S. Fish and Wildlife Service,
Atlanta, GA.
USFWS (United States Fish and Wildlife Service) 1998. Federal and
State Endangered Species Expenditures. Fiscal Year 1995.
Washington, DC.
Vankat, J.L., Major, J., 1977. Vegetation changes in Sequoia National
Park, California. Journal of Biogeography 5, 377±402.
Wilcove, D.S., Rothstein, D., Dubow, J., Phillips, A., Losos, E., 1998.
Quantifying threats to imperiled species in the United States.
Bioscience 48, 607±615.
84 M.W. Schwartz et al. / Biological Conservation 95 (2000) 77±84