DSpace at VNU: Before it had a name: Diagnostic characteristics, geographic distribution, and the conservation of Cupressus tonkinensis (Cupressaceae)

DSpace at VNU: Before it had a name: Diagnostic characteristics, geographic distribution, and the conservation of Cupres...

distribution, and the conservation of Cupressus tonkinensis

(Cupressaceae)

1Lewis B and Dorothy Cullman Program for Molecular Systematics, The New York Botanical Garden, Bronx, NY 10458-5126, USA; e-mail: dlittle@nybg.org

2Royal Botanic Garden, Edinburgh EH3 5LR, Scotland, UK; e-mail: P.Thomas@rbge.ac.uk

3Institute of Ecology and Biological Resources, Vietnam Academy of Natural Science and Technology, 18 Hoang Quoc Viet, Cau Giay, Ha Noi, Việt Nam; e-mail: ntienhiep@hn.vnn.vn

4Center for Plant Conservation, Vietnam Union of Science and Technology Associations

(VUSTA), No 25/32 Lane 191, L ạc Long Quan Rd, Nghia Do, Cau Giay, Ha Noi, Việt Nam

5Hanoi University of Science, 334 Nguyen Trai, Thanh Xuan, Ha Noi, Việt Nam; e-mail: pkeloc@yahoo.com

Abstract In northern Việt Nam the felling of Cupressus for wood and incense has

resulted in the decimation of wild cypress populations Ever since specimens of

Cupre-ssus from Vi ệt Nam were first sent to Paris in 1905, they have been inappropriately

referred to a variety of names including Cupressus funebris, Cu torulosa, and synonyms

of Callitropsis lusitanica This taxonomic disarray was somewhat ameliorated in 1994

when a new species—Cu tonkinensis—was proposed Unfortunately, the morphological

differences between Cu tonkinensis and similar species were not clearly articulated As a

result, Cu tonkinensis has generally been ignored or treated as a synonym of Cu

toru-losa These two species can be distinguished by the presence of dimorphic leaves in Cu.

tonkinensis and their absence in Cu torulosa as well as several continuous morphological

characteristics including the frequency of resin glands (greater in Cu tonkinensis) on

ultimate segments, and the size of the ovulate cone (smaller in Cu tonkinensis) The

natural habitat occupied by Cu tonkinensis in a principal coordinates plot of nine

monthly climate variables and 40 soil variables is remote from other species Cupressus

tonkinensis habitat is statistically distinct from Cu chengiana, Cu gigantea, and Cu.

torulosa, but statistically indistinguishable from the habitats of Cu funebris.

Key Words: Callitropsis, conifers, Cupressus funebris, Cupressus tonkinensis,

Cupressus torulosa, Thông, Vi ệt Nam, Vietnam.

Abstract.

The conifers of Việt Nam have been the

subject of considerable interest over the last

decade The most recent taxonomic account

(Nguyễn & Vidal, 1996 ) described 27

indi-genous species from 15 genera Since then,

six new taxa, including three new species,

have been described (Businsky, 1999 , 2004 ;

Farjon et al., 2002 ; Averyanov et al., 2004 ,

2008 ) and several other genera and species

have been recorded for the first time (Phan &

Nguyễn, 1997 , 1999 ; Nguyễn et al., 2000a ,

2002 , 2004 ; Phan et al., 2002 ) Currently, 32

species from 19 genera are known.

Conifers are at risk throughout Việt Nam

from deforestation and over exploitation of

their valuable timber A recent review of

conifer conservation status in Vi ệt Nam

found that 90% of taxa were nationally

threatened, and 40% were globally threatened

(Nguyễn et al., 2004 ) Three species were

listed as nationally or globally Critically

Endangered, with a fourth from the genus

Cupressus L., listed as Data Deficient

(Crit-ically Endangered) —due to uncertainty about

its identification, native status, and global

distribution Since its initial discovery, the

species of Vietnamese Cupressus have been

referred to a variety of names including

Cupressus funebris Endl., Cu torulosa D.

Don, Cu tonkinensis Silba, and synonyms of

Callitropsis lusitanica (Mill.) D P Little.

Resolution of this taxonomic uncertainty

would warrant a reassessment of Critically

Endangered status due to extremely high

levels of exploitation since the 1950s (Nguyễn

et al., 2004 ) An assessment of Extinct in the

Wild would make it the first conifer species to

be extirpated in modern times.

The first known collection of Vietnamese

Cupressus was made in 1905 under the

auspices of Souère Forestière du Tonkin

(Castellini 169, P) near Sông Hóa, Hoà Lạc

commune, Lũng district, Lạng

province in the northern Việt Nam The

specimen was sterile, with dimorphic ultimate

segment leaves The French botanist Philippe Eberhardt collected a fertile specimen (Eber- hardt 5073, NY; P; Fig 1 ) of the same species growing on calcareous substrate at

“Kai Kinsh” in Lạng province In 1913, another French collector, Auguste Chevalier (Chevalier 29662, P), also collected a Cupressus from a place called “Cai–khinh” (variously spelled “Kai Kung” and “Kai King ”; Chevalier, 1919 , 1944; Hickel,

1930 ) In the early taxonomic accounts of the conifers of Việt Nam (Chevalier, 1919 ; Hickel, 1930 ) the Eberhardt and Chevalier collection localities are treated as identical— occurring at, or near, Kilometer 113 of the Lạng railroad Modern maps of Lạng province include a commune called “Cai Kinh” in Lũng district The center of Cai Kinh is approximately 110 km (by road) from Hà N ˆ:oi Sông Hóa—the locality of the Castellini collection—is within 8.5 km of Cai Kinh Chevalier’s later account ( 1944 ) indi- cates that his collection came from the village of “Quan Lan” in Lạng province.

“Quan Lan” certainly corresponds to Quang Lang on modern maps, in Chi Lăng district, Lạng province, about 20 km to the northeast of Sông Hoá.

Chevalier ( 1919 ) noted that the same species of Cupressus was found in the karst dominated province of Thái Nguyên—a province that adjoins the southwest boundary

of Lạng and the western boundary of Giang During our research we have been unable to locate any Cupressus specimens from Thái Nguyên.

Chevalier’s 1944 account stated that the same species of Cupressus had been collected

at the village of “Bac–Lé” in Giang province by the “Société Forestière” (Souère Forestière du Tonkin) and collected again from the same locality in 1922 by Mignucci (s.n.) Hickel ( 1930 ) simply listed the prov- ince as the locality and cited “Sté forè no 8507” Modern maps of Việt Nam place “Băc

Lê” in Hòa commune, Lũng district,

L ạng province —less than nine

kilo-meters northwest of the Lạng Giang

border Chevalier ( 1944 ) also stated that he

had seen this species of Cupressus in other

localities in northern Việt Nam during his

travels, but did not collect specimens.

Recent wild collections from Lạng have

much more precise collection localities From

Chi L ăng district specimens have been

col-lected at Mạc commune ( 39, HNU),

commune (L ˆ:oc 2004,

P-2005, HNU), townlet ( 2336,

HNU), and Vạn Linh Commune ( et al.

1034, LE, MO) In addition, specimens have

been collected from Liên Nature Reserve

in L ũng (North Vietnam Second Darwin

Expedition [Thomas & Luu] 3, E; Nguyên et al.

HAL11918, E, HN).

According to Chevalier ( 1944 ), the

speci-mens he collected and those collected by the

Souère Forestière du Tonkin were deposited

in the Muséum National d ’Histoire Naturelle

in Paris (P) With the exception of the specimen collected by Castellini, these speci- mens could not be found among the collections of P and probably have been lost (G Aymonin, pers comm.) Silba ( 1998 ) indicates that a Cupressus specimen collected

by Chevalier (Chevalier s.n.) was found at P, but we could not relocate it The collection made by Mignucci has not been relocated.

In total, there are nine wild Cupressus collections known from Việt Nam The local- ities are all within the rugged limestone karst

of the Cai Kinh massif on the eastern side of

River in Lũng and Chi Lăng districts of Lạng province (Fig 2 ) The total area of occupancy is 8 km2(using a 2×

2 km grid) The top ridges of the Cai Kinh massif are ca 300–550 m, rarely reaching

650 m In the Lũng area—near where

FIG 1 Cupressus tonkinensis: A Holotype specimen (Eberhardt 5073, NY; image provided by the DigitalImaging Laboratory at NY) B Isotype specimen (Eberhardt 5073, P)

the Eberhardt specimen was collected—

rainfall is ca 1450 mm, with a five month

dry season from November to March (less

than 50 mm/month) The annual average

temperature is 22.8°C and the average

monthly temperature range is 15.3–28.5°C.

Three of the driest months are also amongst

the coldest—December through February

average less than 17°C (Nguyễn et al.,

evergreen broad-leaved forest Wild Cupressus

is found on higher slopes, on and near ridge

tops—the vegetation there is closed, or

somewhat open, evergreen broad-leaved forest

dominated by Excentrodendron tonkinense (A.

Chev.) H T Chang & R H Miao (Tiliaceae;

Averyanov et al., 2003a , 2003b ).

The sterile specimen collected by Castellini

was not identified by its collector, but

some-time after its arrival at P (19 Jan 1907) it was

annotated (perhaps by Hickel) as Cupressus

torulosa —a species with monomorphic

ulti-mate segment leaves Eberhardt identified his

collections as Cu funebris (Eberhardt 5073,

NY)—a species with dimorphic ultimate

seg-ment leaves After the arrival of one of

Eberhardt’s specimens at P (Sep 1919), it

was annotated by Hickel as Cu torulosa

(Eberhardt 5073, P) Chevalier ( 1919 ) tified the Lạng collections as Cu funebris in his annotated check–list of the forest trees of Tonkin Hickel’s ( 1930 ) anno- tated check –list of Indo–Chinese conifers cited the specimens collected by Chevalier, Eberhardt, and the specimen from L ệ collected by the Souère Forestière du Tonkin

iden-as Cu torulosa—an identification later ted by Chevalier ( 1944 ) Mysteriously, the collection made by Castellini was not cited by either Chevalier ( 1919 ; 1944) or Hickel ( 1930 ).

adop-Later collections of wild Cupressus from

Vi ệt Nam ( 39, HNU; L ˆ:oc 2004,

P-2005, HNU; 2336, HNU; et al.

1034, LE, MO) were also identified as Cu torulosa—presumably using Hickel’s identi- fications as precedent In the account of the gymnosperms for the Flore du Cambodge, du Laos et du Việt Nam (Nguyễn & Vidal,

1996 ), the collections made by Eberhardt, Chevalier, Mignucci, and et al are also cited under Cu torulosa Castellini’s collec- tion was cited as Cu funebris.

In addition to Cupressus torulosa, Cu funebris is also reported from Việt Nam All known specimens of Cu funebris from Vi ệt Nam have been collected from cultivated trees Hickel ( 1930 ) cited a single specimen

of Cu funebris (Capus s.n., P) and noted that this species was considered extremely rare— only occurring at higher elevation (ca.

1000 m) in Tonkin and the south of China Chevalier ( 1944 ) indicated that the Cu funebris specimens from Indo-China that he had seen were cultivated and thereby excluded Cu funebris from his conifer treat- ment Besides the Capus collection (precise locality unknown), Nguyên & Vidal ( 1996 ) listed two additional specimens of Cu funeb- ris (Castellini 169, P [later identified later as

Cu tonkinensis]; P K Lôc s.n., LE [collected from a cultivated tree at Văn, Hà Giang province]) Collections of apparently culti- vated Cu funebris have been made more recently in the karst areas of Hà Giang in northern–most Việt Nam (e.g., Luu s.n., E; P.

K Loc s.n., LE; P K Loc et al HAL8677, HAL11342, HN, LE, MO; Averyanov et al HAL8479, HN, LE, MO) Almost all of these collections are from small trees—probably

FIG 2 Localities of Cu tonkinensis in Lũng

and Chi Lăng districts, Lạng province, Việt Nam

Grid squares are 2 × 2 km The extent of natural

occurrence (204 km2) is outlined in grey Filled stars

represent wild collections Open stars represent cultivated

collections This map is centered on the star in Fig 3

(Albers projection)

less than 20–30 years old Cupressus funebris

is widely cultivated in southern and eastern

China and its historic distribution is unclear

(Farjon, 2005 ).

The literature relating to the habitat and

ecology of wild Cupressus in Việt Nam is

sparse and somewhat confused In Nguyễn &

Vidal’s ( 1996 ) account, they state that it

occurs in Lạng province at altitudes

between 250 –1500 m on semi-arid limestone

hills and is locally associated with

semi-deciduous forests that have a very short

period between leaf shedding and flushing

(i.e., almost evergreen) They also indicate

that Cu torulosa is cultivated in communal

gardens in Lâm province (in southern

Vi ệt Nam; Nguyễn Khôi 82, HN) Other

sources (Vu, 1996 ; Thái, 2000 ; Furey et al.,

2002 ; Nguyễn et al., 2004 ) indicate that wild

Cupressus is restricted to, and sparsely

dis-tributed along, limestone ridges in Lạng

and nearby provinces from 500–700 m.

True Cupressus torulosa is rarely

culti-vated in Việt Nam—to our knowledge it has

never been recorded in secondary vegetation.

In addition, we have not been able to locate

any specimens from areas adjacent to the Cai

Kinh massif of Lạng province, from

other provinces of Việt Nam, or from China

that would indicate widespread cultivation or

naturalization Chevalier ( 1944 ) clearly

dis-tinguished between cultivated and wild

speci-mens and obviously believed that the

collections from Vi ệt Nam he identified as

Cu torulosa were of wild origin The

application of Cupressus torulosa to

Viet-namese Cupressus specimens from the Lạng

area has persisted despite the extreme

geographic and morphological disjunction

between the Himalayan (India, Nepal, and

Pakistan) and the supposed Vietnamese

pop-ulations (compare squares versus the star in

Fig 3 ).

In 1994, Silba described Cupressus

tonki-nensis, designating “Eberhardt 173” (NY) as

the holotype (Fig 1A ) Possibly because the

handwriting on the label is ambiguous, there

is confusion in the literature concerning the

collection number of the Eberhardt

speci-mens: Hickel ( 1930 ), Chevalier ( 1944 ),

Farjon ( 2005 ), and Rushforth ( 2007 ) cite

“5073” whereas Silba cites either “173”

(Silba, 1994 , 1998) or “1073” (Silba, 2005 ) Silba ( 1994 , 1998, 2005) identified additional presumably wild specimens collected from Guizhou China as Cu tonkinensis (Cheng

2362, A; Steward et al., 10, A, L; Tsiang

8004, 5123, A) as well as cultivated mens from Yunnan, China (Chu 51411,

speci-51421, KUN, NY) The description was self–published and is not widely available: it was unavailable to Nguyễn & Vidal ( 1996 ) when they were preparing their account In his description, Silba ( 1994 ) distinguished between Cu tonkinensis and Cu chengiana

S Y Hu using the arrangement and relative thickness of the branches Silba also distin- guished between Cu tonkinensis and Cu lindleyi Klotzsch ex Endl (= Callitropsis lusitanica) using the shape of the leaf apex Silba ( 2005 ) later treated Cu tonkinensis as a subspecies of Cu funebris Although Silba did not clearly articulate the differences between Cu tonkinensis and Cu torulosa, other researchers (Rushforth et al., 2003 ; Rushforth, 2007 ) noted the contrasting dimor- phic leaves of Cu tonkinensis and the monomorphic leaves of Cu torulosa Farjon ( 2005 ) dismissed Silba’s work and synonymized Cupressus tonkinensis with Cu torulosa—perhaps because the dimorphic leaf character was not obvious in the literature and usually cannot be observed on herbarium specimens without some dissection (dimor- phic-leaved species are often mounted such that all of the facial leaves are parallel to the herbarium sheet while at the same time all of the lateral leaves are perpendicular making the contrast between the two dif ficult to detect without some manipulation) Farjon did not explain the discrepancy in leaf type between the two species—simply describing the leaves of Cu torulosa s.l (including Cu gigantea and Cu tonkinensis) as monomor- phic Farjon explained the geographic dis- junct as the result of human introduction of

Cu torulosa at “an unknown time”, followed

by naturalization “on semi-arid slopes with secondary vegetation, mainly on limestone”.

In his monograph, Farjon restricts Cu ulosa to the dry inner valleys and semi-arid high mountain environments of the eastern Himalayas at altitudes between 1560–3670 m where rainfall is not more than 300 mm; its

tor-presence in Yunnan is also attributed to

human introduction at an unknown time.

The distinctiveness of Cupressus

tonkinen-sis from Cu torulosa (and other species) is

strongly supported by an analysis of DNA

sequence data combined with anatomy,

bio-chemistry, and morphology (Fig 4 ; Little,

2006 ) Cupressus tonkinensis is included

within a clade of Asian species with

dimor-phic ultimate segment leaves (Cu chengiana,

Cu funebris, and Cu jiangensis N Zhao)

while Cu torulosa is nested in a clade of

Asian species with monomorphic (Cu.

duclouxiana Hickel in A Camus and Cu.

gigantea W C Cheng & L K Fu) or weakly

dimorphic ultimate segment leaves (Cu

aus-trotibetica Silba and Cu cashmeriana Royle

ex Carrière) The clade containing Cu.

tonkinensis is sister to the clade containing

Cu torulosa plus a clade of Eurasian and

north African species with weakly dimorphic

ultimate segment leaves (Cu atlantica Gaussen,

Cu dupreziana A Camus, and Cu

sempervi-rens L.).

Regardless of its identity, the Cupressus in

Việt Nam has always been valued for its fine

timber and resin; consequently it has always

been heavily exploited One indication of this

can be seen in the extraordinary cost of a cubic

meter of wood: in 1987 it was 3 million ,

by 1997 it had increased 10 times to 30 million

During the same period other high value timbers and resins such as Fokienia hodginsii

A Henry & H H Thomas (locally known as

“Pomu”) went from 700,000 to 5 million per cubic meter After 1997, no commer- cial quantities of Cupressus were available and

it virtually disappeared from the market (Bay,

2003 ) Restrictions on its harvest were duced as early as 1958 (Thái, 2000 ), but little was done to stop the exploitation Between

intro-1986 and 1989 a 10,000 ha nature reserve was established in Liên to protect populations

of musk deer, limestone forests, and the remaining populations of Cu “torulosa” (Furey et al., 2002 ) Cupressus “torulosa” was listed in the Group IA protected species under Vietnamese law: this prohibited any further exploitation or use (Anonymous,

2006 ) It was listed in Việt Nam’s Red Data book as Critically Endangered (CR A1a,d; Anonymous, 2007 ) In the late 1990s a small plantation was established by the Forest Sci- ence Institute at the edge of the Liên Reserve By 2004, what was thought to be the last known natural tree in the reserve was destroyed in a fire (Nguyễn et al., 2004 ) More recently (October 2008), a single tree was located in another part of the massif Apart from this, no other wild Cupressus trees are known to occur in Việt Nam This species is also represented by about 20 trees remaining in

FIG 3 The distribution of wild Cupressus chengiana (triangles), Cu funebris (circles), Cu gigantea (diamonds),

Cu tonkinensis (star; with arrow), and Cu torulosa (squares) Cupressus jiangensis is known only from cultivationand therefore excluded from thisfigure Callitropsis lusitanica is native to the New World (distribution not shown;Mercator projection)

the plantation, several cultivated trees growing

in the surrounding villages, and a clonal

collection held by organizations within the

Ministry of Agriculture and Forestry The

origin, and the identity, of some of these trees

are uncertain (Nguyễn & Thomas, 2004 ;

Nguyễn et al., 2004 ).

This paper strives to clearly articulate and

test the morphological and ecological

differ-ences between Cupressus tonkinensis and the

species that could potentially be confused

with it This information should help to

identify additional wild plants, should they

be found, and also identify those held in ex

situ collections.

Methods and materials

Published climate and soil data were used

in combination with georeferenced herbarium

specimens to test for ecological differences

between Cupressus tonkinensis and related

species Continuous and discrete variation in

morphological characteristics was used to test

for morphological differentiation.

The species sampled include those found in

the least inclusive clade containing Cupressus

tonkinensis (Fig 4 ) and the species that have

historically been confused with Cu

tonkinen-sis or associated taxonomically (Callitroptonkinen-sis

lusitanica, Cu chengiana, Cu funebris, Cu gigantea, Cu jiangensis, Cu tonkinensis, and

Cu torulosa) A complete list of specimens examined can be found in Appendix 1

If available, label coordinates from herbarium specimens were used directly Specimens with- out coordinate information were georeferenced using data downloaded from GNS ( http://

Georeferencing used the most precise locality provided on the label that could be found in the databases All specimens were considered wild unless the label explicitly indicated otherwise Maps were constructed using GMT version 4.1.1 (Wessel & Smith, 2006 ).

For wild specimen localities in Asia, nine monthly climate variables (cloud cover, diur- nal temperature range, minimum temperature, mean temperature, maximum temperature, ground-frost frequency, precipitation, vapor pressure, and wet day frequency) were extracted from the 10-year monthly-mean (1981 –1990) data set of New et al ( 2000 ; downloaded from http://www.daac.ornl.gov ).

As published, the climate data are lated and have a gridded resolution of 0.5° Forty soil variables (for each soil profile: soil depth to a physically limiting layer and available water capacity; separately for top- soil and subsoil: aluminum saturation, base saturation, total extractable bases, cation exchange capacity, cation exchange capacity

interpo-of the clay fraction, exchangeable sodium, pH

in water, electrical conductivity of a saturated paste, bulk density, calcium carbonate con- tent, gypsum content, organic carbon, total nitrogen, organic carbon/nitrogen ratio, clay content, sand content, silt content, gravel content, and total porosity) were extracted for wild specimen localities from the unbanned data set of Batjes ( 2005 ; down- loaded from http://www.isric.org ) As down- loaded, the 0.5° gridded unbinned soil dataset included the variation among soil units within each grid cell—this variation was reduced to

a single value per variable per grid cell by calculating a mean weighted by the relative area within the grid cell occupied by each soil unit Because the number of herbarium speci- mens from a particular region does not necessarily reflect population size, the data were filtered for each species prior to analysis

to remove duplicate grid squares.

FIG 4 A phylogenetic hypothesis for the genus

Cupressus (sensu Little, 2006) based on data from

anatomy, biochemistry, morphology, and DNA sequence

(matK, NEEDLY intron 2, nrITS, rbcL, and trnL) redrawn

from Little (2006) Numbers below branches indicate

jackknife values The number of terminals contained

within each condensed clade is indicated at the clade

base

For scanning electron microscopy,

field-col-lected specimens were fixed in FAA (50% or

70% ethanol, 5% acetic acid, and 5% formalin).

After a minimum of 48 h of fixation, samples

were transferred to a solution of 50% or 70%

ethanol and 10% glycerin for storage Samples

were dehydrated in an ethanol series (70%,

95%, 100%, 100%, 100%) and then transferred

to a 1:1 ethanol:acetone solution Expulsion of

ethanol was accomplished with three changes of

pure acetone Specimens were transitioned from

acetone to liquid carbon dioxide and

critical-point dried in a Denton DCP-1 (Denton

Vacuum, Moorestown, NJ) Specimens were

mounted on aluminum stubs with carbon tape

and sputter coated with gold/palladium for 4

minutes using a Hummer 6.2 sputtering system

(Anatech, Union City, CA) Observations were

made with a JSM-5410LV (JEOL, Tokyo) SEM

using an accelerating voltage of 15 kV.

Measurements of continuously distributed

morphological characteristics and scores for

discrete characters were obtained from

herba-rium specimens Characters and scoring

tech-niques are detailed in Appendix 2 We

attempted to gather three measurements for

each continuous characteristic from each

speci-men examined, but the amount of suitable

material often limited the number of

measure-ments Prior to statistical analysis

intra-individ-ual continuous characteristic variation was

eliminated by using the median value for each

characteristic from each specimen.

Statistical tests and visualizations were

conducted with R version 2.6.2 (R

develop-ment core team, 2008 ) and the R packages

LabDSV version 1.3-1 (Roberts, 2008 ),

pgirmess version 1.3.6 (Giraudoux, 2008 ),

and vegan version 1.11-4 (Oksanen et al.,

2008 ) The variation in the climate and soil

data was visualized by principal coordinates

analysis (Gower, 1966 ) of Manhattan

dis-tances calculated from mean-centered and

root-mean-squared scaled data (scale, dist,

and pco functions) A non-parametric

multi-ple-comparison test based on the

Kruskal-Wallis (Kruskal & Kruskal-Wallis, 1952 ; Siegel &

Castellan, 1988 ) rank sum test (kruskalmc

function) was used to determine if the

species were distinct at the p = 0.05 level for

each of the 148 soil and climate variables.

The output was corrected for multiple tests

using the method of Benjamini and Hochberg

( 1995 ; p.adjust function) The variation in continuous morphological characteristics was analyzed using the multiple-comparisons method described above.

Results For the nine monthly climate variables and

40 soil variables, the two principal nates axes together account for 71.5% of the variation in the original data—PCO 0 accounts for 61.7% and PCO 1 accounts for 9.8% (Fig 5 ) The habitat occupied by Cupressus tonkinensis in the principal coor- dinates plot is remote from that of Cu torulosa (Fig 5 ) The area between Cu tonkinensis and Cu torulosa is filled by habitats occupied by Cu chengiana, Cu funebris, and Cu gigantea Cupressus funebris is geographically widespread and occupies a wide range of habitats —as a result

coordi-it approaches or envelopes every other species represented in the plot On a per character basis, the habitats of Cu funebris are statisti- cally distinct from those occupied by all other species except Cu tonkinensis (Table I ) The habitat of Cu tonkinensis is statistically distinct from that of Cu chengiana, Cu torulosa, and Cu gigantea The habitats of

Cu tonkinensis and Cu torulosa differ cally for 71 of the 148 soil and climate variables Gradients of climate and soil varia- tion generally run parallel to PCO 1 Among

statisti-t h e statisti-t o p 2 5 % b e s statisti-t fitting variables (D2=0.9570–0.9901) all but three (subsoil

pH in water, subsoil electrical conductivity of

a saturated paste, and subsoil cation exchange capacity of the clay fraction; Fig 5D ) are climate variables —mostly relating to the weather in the months of May through October (64.8% of the top ranked variables) The top 25% best fitting variables include minimum temperature for all twelve months (e.g., Fig 5A ), nine months worth of vapor pressure variables (e.g., Fig 5B ), and seven months worth of mean temperature variables Discrete morphological characters can be used to distinguish Cupressus tonkinensis from Callitropsis lusitanica, Cu funebris, Cu gigantea, Cu jiangensis, and Cu torulosa (Tables I and II ) Of particular interest are the distribution of dimorphic leaves (Figs 6 – 8 ; Table II , characters 34 and 37) and mucronate

leaf apices (Fig 8 ; Table II , character 39) In

addition, continuous morphological variation

can be used to statistically distinguish Cu.

tonkinensis from Cu chengiana, Cu funebris,

Cu gigantea, and Cu torulosa (Tables I and III ,

characters 2, 5, 10, 11, 15, 23, 24, and 25).

Across all taxa, two of the five most statistically

discriminatory characteristics (Table III ) are

vegetative: ultimate segment length (character

10) and ultimate segment leaf length (character

14) The remaining three most discriminatory

characters depict features of the ovulate cone: length (character 23), width (character 24), and pairs of scales (character 25).

Discussion Historically Cupressus tonkinensis has been confused with Cu torulosa (Hickel, 1930 ; Chevalier, 1944 ; Nguyễn & Vidal, 1996 ; Farjon, 2005 ) despite the extreme geographic separation (Fig 3 ) and marked difference in

FIG 5 Principal coordinates analysis of nine monthly climate variables and forty soil variables for wild specimenlocalities for Cupressus chengiana (triangles), Cu funebris (circles), Cu gigantea (diamonds), Cu tonkinensis (stars),and Cu torulosa (squares) Wild populations of Cu jiangensis are unknown and no native Asian populations ofCallitropsis lusitanica are known therefore these species were excluded from this analysis Variation in four arbitraryselected variables are shown asfitted surfaces A Minimum September temperature (D2=0.9894) B September vaporpressure (D2=0.9786) C September ground frost frequency (D2= 0.9581) D Subsoil pH in water (D2=0.9703)

habitat preference (Fig 5 ; Table I ) However,

these species are easily distinguished by the

presence of dimorphic leaves in Cu tonkinensis

and their absence in Cu torulosa (Tables I and

II , characters 34 and 37; Figs 6E–G , 7E–G ) as

well as many continuous characteristics

includ-ing the frequency of resin glands on ultimate

segments (Table III , character 15) and the size

of the ovulate cone (Table III , characters 23, 24,

and 25).

Cultivated Callitropsis lusitanica has

also been confused with Cupressus

tonki-nensis The presence of dimorphic leaves in

Cu tonkinensis is the most diagnostic

characteristic (Figs 6E–F , 7H , and 8H ;

characters 34 and 37), but penultimate

branches arranged on a single plane

(char-acter 32) and obtuse leaf apices (char(char-acter

38; Fig 8F and 8D ; Silba, 1994 ) are also useful (Table II ).

Silba ( 1994 ) used stem diameter (characters

2 and 11) to distinguish between Cupressus tonkinensis and Cu chengiana—the differ- ences in these characters are statistically sig- nificant (Table III ) Additional significant continuous characters include ultimate branch segment length (Table III , character 10), the length of the ovulate cone (Table III , character 23), and the number of ovulate cone scales (Table III , character 25).

Cupressus tonkinensis is perhaps most easily confused with Cu funebris These two species can be distinguished by the absence of a mucronate tip on the apex of the ultimate segment lateral leaves in Cu tonkinensis (Fig 8F ) and the presence of such

TABLEI

PERCENTAGE OF PAIR-WISE NON-PARAMETRIC TESTS RESULTING IN SIGNIFICANT DIFFERENCES AT P=0.05 (CORRECTED FOR MULTIPLE –COMPARISONS AND TESTS) FOR CLIMATE AND SOIL VARIABLES (BELOW THE DIAGONAL) AND CONTINUOUS MORPHOLOGICAL CHARACTERS (ABOVE THE DIAGONAL) WHEN APPLICABLE DIFFERING DISCRETE MORPHOLOGICAL CHARAC- TERS ARE GIVEN PARENTHETICALLY WILD POPULATIONS OF CUPRESSUS JIANGENSIS ARE UNKNOWN AND NO NATIVEASIAN POPULATIONS OF CALLITROPSIS LUSITANICA ARE KNOWN, THEREFORE STATISTICAL TESTS OF CLIMATE AND SOIL VARIABLES ARE

NOT APPLICABLE (NA) FOR THESE SPECIES

Ca lusitanica Cu chengiana Cu funebris Cu gigantea Cu jiangensis Cu tonkinensis Cu torulosa

Ca lusitanica — 0% (34) 25% (32,

34, 37)

23% (36, 38,44)

POLYMORPHISMS ARE ENCLOSED IN BRACKETS VARIATION IN ULTIMATE SEGMENT LEAF FORM(CHARACTER37)IS ILLUSTRATED INFIGS.6,7,AND8 VARIATION IN ULTIMATE SEGMENT LEAF APICES(CHARACTERS38AND39)IS ILLUSTRATED

a tip on the ultimate segment lateral leaves of

Cu funebris (Fig 8B ).

Although the habitat occupied by

Cupres-sus tonkinensis is distant from the habitats

occupied by other species in the principal coordinates plot (Fig 5 ), the habitat is not statistically distinct in all cases (Table I ) All

of the verified Cu tonkinensis collections

FIG 6 Ultimate segments with attached leaves A, B Cupressus funebris (Little et al 862, BH, CDBI) A Facialleaves B Lateral leaves C, D Cupressus gigantean (anon 2002 Sep 11, BH) C Leaves D Rotated 90° from thatshown in C—indicating that the leaves are not dimorphic E, F Cupressus tonkinensis (NY-HN 110, HN, NY) E.Facial leaves F Lateral leaves G, H Cupressus torulosa (Little 632, BH, TUCH) G Leaves H Rotated 90° fromthat shown in G—indicating that the leaves are not dimorphic I, J Cupressus jiangensis (Little & Sun 869, BH,CDBI) I Facial leaves J Lateral leaves K, L Cupressus chengiana (Little et al 862, BH, CDBI) K Facial leaves

L Lateral leaves

are from an area less than 0.5° square,

therefore Cu tonkinensis is represented in

the habitat analysis by just two data points

whereas other species are represented by 5–

85 data points each The Kruskal-Wallis type test (Kruskal & Wallis, 1952 ; Siegel & Castellan, 1988 ) used for the analysis is robust to unequal variance and sample size,

FIG 7 A–G, I–L Ultimate segment leaf margins at the mid point of the leaf, adaxial view A, B Cupressusfunebris (Little & Chen 847, BH, KUN) A Facial leaf B Lateral leaf C Cupressus gigantea, leaf (anon 2002 Sep

11, BH) D Callitropsis lusitanica, leaf (Little & Ochoterena 795, BH, MEXU) E, F Cupressus tonkinensis (NY-HN

110, HN, NY) E Facial leaf F Lateral leaf G Cupressus torulosa, leaf (Little 632, BH, TUCH) H Callitropsislusitanica, an ultimate segment with attached leaves (Little & Ochoterena 795, BH, MEXU) I, J Cupressusjiangensis (Little & Sun 869, BH, CDBI) I Facial leaf J Lateral leaf K, L Cupressus chengiana (Little et al 862,

BH, CDBI) K Facial leaf L Lateral leaf

but results obtained using such small sample

sizes should be treated with caution

Increas-ing the sample size for Cu tonkinensis in the

habitat analysis would likely increase the

number of significant comparisons, but this

is not possible given the biological reality—

Cu tonkinensis is a narrow endemic whereas the other species range from wide-

FIG 8 A–G, I–L Ultimate segment leaf margins and apices, adaxial view A, B Cupressus funebris (Little &Chen 847, BH, KUN) A Facial leaf B Lateral leaf C Cupressus gigantean, leaf (anon 2002 Sep 11, BH) D.Callitropsis lusitanica, leaf (Little & Ochoterena 795, BH, MEXU) E, F Cupressus tonkinensis (NY-HN 110, HN,NY) E Facial leaf F Lateral leaf G Cupressus torulosa, leaf (Little 632, BH, TUCH) H Callitropsis lusitanica, anultimate segment with attached leaves, rotated 90° from that shown in Fig.7H—indicating that the leaves are notdimorphic (Little & Ochoterena 795, BH, MEXU) I, J Cupressus jiangensis (Little & Sun 869, BH, CDBI) I Facialleaf J Lateral leaf K, L Cupressus chengiana (Little et al 862, BH, CDBI) K Facial leaf L Lateral leaf