* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Morphometric analyses of mixed Dactylorhiza colonies (Orchidaceae)
Ecology of Banksia wikipedia , lookup
Plant defense against herbivory wikipedia , lookup
History of herbalism wikipedia , lookup
Plant breeding wikipedia , lookup
Plant physiology wikipedia , lookup
History of botany wikipedia , lookup
Plant morphology wikipedia , lookup
Evolutionary history of plants wikipedia , lookup
Plant use of endophytic fungi in defense wikipedia , lookup
Flowering plant wikipedia , lookup
Plant ecology wikipedia , lookup
Plant evolutionary developmental biology wikipedia , lookup
Plant reproduction wikipedia , lookup
Ornamental bulbous plant wikipedia , lookup
Hybrid (biology) wikipedia , lookup
BotaizicalJournuloJfthe Linneun So&@ (1998), 128: 385 401. With 5 figures Artick ID:bt980203 Morphometric analyses of mixed Dactylorhiza colonies (Orchidaceae) on industrial Gaste sites in England PETER J. A. SHAW School of Lzji Sciences, Roehumpton Institute, Whitelands College, West Hill, London S W 1 5 3 S N Received October 1997; acceptedf w publication3;& 1998 Morphometric data were collected from Dac&lorhiza growing on two types of industrial waste (pulverized fuel ash or PFA, and Leblanc process waste) during the summer of 1997. Three species grew an PFA (0. fuchsii, D . incamata, D. praetemksa). The same species plus D. pu7purella grew on the Leblanc site, although on both substrates the majority of plants failed to correspond precisely with published descriptions, introducing an element of subjectivity into the field identifications. Principal Components Analysis and Detrended Correspondence Analysis ordinations confirmed that textbook species descriptions corresponded to extremes of multivariate space. Cluster Analysis failed to produce a useful resolution of the data. Discriminant Functions Analysis initially gave the misleading result that any plant with spotted leaves was D.fuchsii, but produced useful results after leaf spotting was removed from the analysis. O n PFA sites hybrids appeared to be mainly D. praetemzissa x D. j%chsii (= D. grandis) or D. praetermissa x D. incamata (= D. wintoni). The identity of hybrids on the Leblanc site was unclear, perhaps reflecting the greater age of this site which may have allowed extensive introgression. 0 1998 The Idnnean Society of London ADDITIONAL KEY WORDS:-colonization waste. - conservation - industrial . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386 388 388 388 389 390 390 392 398 400 400 ~ hybridization CONTENTS Introduction . . . . . . . . . . . . . . Methods . . . . . . . . . . . . . . . Site descriptions and plant selection . . . Morphometric measurements . . . . . Characters measured . . . . . . . . Results . . . . . . . . . . . . . . . Preliminary data description . . . . . . Multivariate analyses . . . . . . . . Discussion . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . References . . . . . . . . . . . . . . 0024-4074/98/120385 + 17 $30.00/0 385 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 1998 The Linnean Society of London Sel era1 hea\s) industrial processes generate such large volumes of solid wastes that their disposal sites become locally important landscape features (Gemmell, 1977). Such derelict land has a poor public image, but many recolonized industrial site5 contain locally important populations of eutrophication-intolerant plants 'Box, 1993; Sham, 1994). It is well known that industrial waste sites frequently support largc populations of orchids, especially where the waste is calcareous (Lee 8r Grcenuood. 1976; Greenkvood & Gcmmell, 1978). At most such sites Daciylorliizn Necker ex Nevski species are the dominant orchids, though Gjmnadenia ionottwa (L.) R.Br. subsp. conop\ea. .hacamptr~ ~ ~ ~ r a m i d a(L.) l i s L.C.M.Rich. and Epzpactic spp. also occur. Xomenclature follows Stace (1 997). From the virwpoint of conservation, the presence of orchids on industrial sites is useful since these plants hake a high public profile and hence have been uscd as a justification for prcxenting site de\ elopment or t o force a rescue-relocation scheme (Shaw, 1994, 1998). However, problems arise lvhen trying to name the species, especially of Dacplorhz,-n, due to the M ell documented case of hybridization between taxa in this genus (Bateman 8r Denholm. 1983, 1985, 1989a). Stace (1997: 974) suggests that a representatixe sample of 5- 10 non-extreme plants should be examined I d o r e allocating a name to n Dac{)*lor/izzapopulation, and that some individuals ma) defy satisfactor) identification. O n e extrcme rcsponse to this variability is shoivn b) the ivarden of Nob End Leblanc waste site, who now refuses to attempt to identif) the thousands of Dacplorhzza on his reserve below gencric lcvel (W. Halton, perc. comm., June 1997). 'Ydcock, Gorton 8r hlorries (1983) attempted to classify the Dactylorhzza growing on industrial sites in hlanchester (including one re-examined in this study). They recorded DacQlothiza incarnata (L.) Soo, D. praeteimtssa (Druce) Soo, D.purpurella (T. & T.-L Stevenson) Sob and D.fuchsii (Druce) Sob. They explicitly avoided measuring the many plants suspected of being hybrids, and restricted their analyses to pol) graphs and bar charts. These limitations are unfortunate, since an adequate description of morphological \ ariation within such populations requires multik ariate anal! ses encompassing putative hybrids as well as standard plants (Bateman & Denholm, 1989a,b; lVilson, 1992). An example ofthis approach i % a 5 provided by Xndersson (1994), who used multivariate analyses to examine the 5tructure of Dnctylorhiza traunsteinui (Saut. ex Rchb.) Sob populations in SItcden. She concludcd that plants lvhich some orchidologists would allocate to four different species in fact represent \.ariation M ithin one heterogenous species. Hedrt.n i 1996a.b) used alloLc).memarkers to explore lineages within DacQlorhira, reporting that I). zncainata and D. fuchsii showed long evolutionary separation, but that D.p m e t c i m i m and D. purpurella appeared to be allotetraploids arising more reccntl) from hybridization between D.zncarnata and D.fuchcii. The aim of the work described here was to transcend the labelling of individual I)nc@lorhzza plants, and to defcribe the fbll extent of morphometric variability within geographicall) separated populations on different industrial wastes. Both univariate and multikariate analyses are used to examine whether any sharp boundaries or clear clusters can be found that could correspond to recognized species distinctions. .Uthough there have been many multi\,ariate analyses of the morphometrics of the genus Dactdorhzza (e.g. Bateman & Denholm, 1983, 1985, 1989a; Rndersson, 1994, MORPH ()METRICS OF DA CTYLORHf<4 387 1995)) this is the first published work using such techniques to explore variation within hybrid swarms specifically on industrial waste sites. Two different industrial wastes were examined, PFA and Leblanc process waste. PFA is the fine ash resulting from burning coal in modern power stations. It is a well-studied material that supports clearly defined vegetational changes during its colonization (Shaw, 1992).After the soluble salts have weathered out (typically 5- 10 years), a legume-calcicole sward develops that often supports a few orchids. Birch/ willow scrub develops and closes canopy after 20-40 years, but the remaining open glades often contain dense hybrid swarms of Dac@ylorhiza (Shaw, 1994). Curiously, other orchid genera tend to be absent. The technology for burning pulverized coal dates from the 1950s, so no PFA sites are older than c. 1955. Leblanc waste is a relic of the Leblanc chemical procedure that was widely used to manufacture bleach, sodium carbonate (washing soda), hydrochloric acid and other chemicals (Mathews, 1978a,b). The production of one ton of sodium carbonate generated approximately two tons of waste, consisting of calcium sulphide mixed with lime, carbon and unburned sulphur. Virtually no Leblanc waste has been produced since 1900 so very few Leblanc sites now survive, all in the industrial north-west. At the Leblanc site studied it is believed that the surface of the waste was subsequently removed for extraction of unburned sulphur by the Mond process, before being redumped around 1890. Over the intervening 100 years the waste weathered down to an infertile gypsum/lime mixture with extremely low levels of nitrogen and phosphorus (Burrows, 1995). It was decided to apply several complementary lines of analysis to the data, to explore the stability of analytical conclusions. In addition to standard univariate analyses, four different multivariate techniques were applied. Two ordination algorithms were used: Principal Components Analysis (PCA) and Detrended Correspondence Analysis (DCA, often erroneously called DECORANA after the computer program which performs the analysis). PCA is the most mathematically natural ordination technique (Gauch 1982; Digby & Kempton 1987) and is closely related to Principal Coordinates Analysis, used by Bateman & Denholm (1983, 1985, 1989a). DCA was originally devised to simplify the interpretation of community data (Hill & Gauch, 1980) and is probably the ordination technique most widely used by ecologists, but has also been recommended as a taxonomic tool (e.g. Parnell & Waldren, 1996). Cluster Analysis is routinely used both in taxonomy and community ecology to classify individuals on the basis of multivariate attributes, although the wide choice of clustering algorithms (Kent & Coker, 1992) and the huge number of ways that any given dendrogram can be portrayed combine to make interpretation problematic. Discriminant Function Analysis, otherwise known as Canonical Variates Analysis (CVA: Gittins, 1985)) requires that at least some members of a sampled population can be assigned to pre-defined categories, and uses the multivariate attributes of these individuals to produce linear functions by which the group membership of additional individuals can be estimated. As such it is a logical choice when attempting to classify problematic individuals. Jeffers (1996) applied these four multivariate analyses to a dataset describing variation in European elms (along with logistic regression and a genetic algorithm), and found the diversity of analyses to be helpful in giving complementary insights into the data structure. TABLE 1. Site dcscriptions .Sik 1. Chrshunt ,gravel pit. [1'L3i lOY2). vc18. PFA froin Britnsdown power station was deposited into a gravcl pit 11) Iiargc in c. 1956. 11 notv dcnse l)irch/\villo\v woods, with orchids in open glades along with a diverse flora including rls/fl- ~ n / j p i i sIfilld., &a.cininc./iin nut~/miilarinI.. and grasses. A moist site, soil pH 7.4. ,Sik 2. Rnrking ash lagoon. "&658'25), vr18. A PF.1 lagoon surrounded by 3 r n high Ilunds adjacrnt t o the '~Ihamrs.c~~ntainiiig I'FA dumped from Barking power station in c. 1965. Plants were moved from here to Nottingharn xs a rescue relocation schemc In E'ehruar). 1997, and biometric data were collected in thrir new site ( i n a new PFA dump at SK456:31Y as plants were establishing. Donor site was a wet lagoon of pH 7.2, whereas the rrcipient sitv \\.as drier and morc alkalinc [pH 8.0). S i k 3. I h a x power station. (SE(i55280). vc6 I . A srries of six PFNLgypsum mounds (each 5 ni x 5 m x 3 m high) wcrc waI&shcd in 1987. I n 1991 soit rraped from C:lirsliuiit arid other orchid-bearing PFA sites, and a thin Ia)cr .c. 0.1 mm') introduced a s an inoculum sourcc. The first three years saw establishment of halophytes and Ityimics (Sliaw. 1996).and orchids have h v e r r d in damp areas at mound haws ever) year since 1995. Substrate pH rnngc.; 8.3 ;piire PFA) to 7.7 !~]isi~ni-coiitaiiiin~ mounds). 4.Carrington po\z-rI wtion, 793 I), vc59. PFA dumped in c. 1975 has formed a marshy commun ?ha lat@/in I,.. Phrapiik ciro/ra/ '.) Trin. cx Strud., and S d t - scrub. Opcn areas near by support dens o C Nlorkr/oninpt~~o-folio~~ [I..) Huds. a n d other annuals. Soil pH 7.5. .Qt, 5. Xrih End. K o l t o n . !S1)737074). vc.59. j\n I comprising >5 ha of Ixblanc waste dating from c. 1900, c.ontaining Sn/i.r scrub. thousands of flor!vforh& nnd f~rrn?indpnioronopscn along and other calricoles. A dry sitc, c l r \ a t d 'LO m ak)ove the riwrs Croal and Invell. Soil p H 7.6. AIKI'HODS Site descriptions and plant selection All sites visited consisted of sizeable areas (>1 ha) of alkaline industrial wastes (Table 1). Data were collected from four PFA dumps, of which the most intensively studied was Cheshunt sgravel pit in the Lee Valley of Hertfordshire, where a boardwalk has been erected to allow public access to a dense stand of Dacglorhzra. One PFA site contained plants which had recently been translocated as part of a rescue-relocation scheme, which could have affected their development. Lehlanc data were collected from the largest and best known surviving Leblanc site, 'Nob End' in Bolton, a Site of Special Scientific Interest where a 10 m thickness of Leblanc waste has been left largely unmanaged since Victorian times. At each site, orchids were systematically sampled from within randomly selected 3 m x 5 m areas during June 1997. A provisional identification was attempted for each plant, based on criteria given in Rose (1 98 l), Clapham, Tutin & Moore (1 989) and Stace (1 997). Remarkably few plants matched precisely all the key characteristics for any given species, and the identification process involved a subjective judgement about the de<greeof allowable variation. Plants which could not satisfactorily be assigned to a species in the field were recorded as putative hybrids. It is important to note that, in the context of this study, the label 'hybrid' denotes only that a decision was made in the field that a plant could not be identified with confidence, not that its features were intermediate between two standard species. .\ forphometric measurements There is a balance in any morphometric study between the number of characters measured, the number of individuals measured in any population and number of MORPHOMETRICS OF DACTYLORHIZA 389 populations studied (Bateman & Denholm, 1989b). In this study the number of characters measured was kept low (23, compared with 52 used by Bateman & Denholm, 1983, 1985, 1989a) in order to allow sampling of a greater range of sites within one season. All characters were measured in situ without damaging the plants. Consequently, floral dimensions are not comparable with those of Bateman & Denholm (1983, 1985, 1989a), who worked on preserved flowers. Floral measurements were recorded to 0.1 mm using a Sylvac digital micrometer, and were replicated on two lowermost flowers because between-flower variation was evident in many plants. Other measurements were recorded to an accuracy of 1 mm using a ruler. Stem solidity was tested by non-destructive fingertip squeezing of the stem base. Characters measured Non$oral measurements overall height, mm number of flowers stem solid/hollow length of first leaf, mm length to widest point of the same leaf presence/absence of leaf spots presence/absence of a hooded leaf tip Floral measurements distance from stem to tip of bract, mm distance from stem to tip of the corresponding flower, mm distance from neck of spur to tip of labellum, mm depth of labellar sinus, mm labellar width (frontal view), mm labellar depth (lateral view), mm spur length, mm Ground colour (red, purple or white; and whether dark or light) Lip pattern (spots only, spots +lines, or a solid central mass of colour). A lipfold angle was calculated as arctanpip depth/(lip width/2)]. Prior to multivariate analysis all floral data were converted to one mean value per plant, to remove within-plant variation. Ground colour was recoded as two variables: PURPLE (0 if red or white, 1 if pale purple, 2 if deep purple), and RED (0 if purple or white, 1 if pale red, 2 if deep red). Two lip pattern variables were included in multivariate analyses: lines @resent/absent) and hyperchromic @resence/absence of a large solid central blotch). Within-plant variation between floral characters (which were measured on two replicate flowers per plant) was examined by a paired samples t-test. Significance of differences between substrata and between species were tested by one-way ANOVA where the data approximated to normality, otherwise the Kruskal-Wallis test was used (Steel & Torrie, 1982: 544). Data were ordinated by PCA (using the Pearson correlation matrix) and Detrended Correspondence Analysis (Hill & Gauch, 1980). Several different Cluster Analysis algorithms were applied, and Discriminant Function Analysis (otherwise known as Canonical Variates Analysis) was applied to all plants P J. .I SH.\\\ 1” I ’I:.u~LF.2. Formal allocation of plants t o species hased 011 the kcys provided 1)- Rose (3981),Clapham, ‘I’utin & lloore 1989:. and Stace (19‘37). The total numl~crof plants allocated to each specics by each kcy is gi\.c.ii, along \\-itti the plants \clrich had one or morr characteristics that mis-matched any part ~ key 143 I43 bawd on the identities allocated in the field to named (i.e. apparently non-hybrid) plants. hlost calculations \$ereperformed by ‘SPSS for Windows’, though the cluster aiial)ses were run on LlVSP 2.0 (Ko\.ach, 1990) and DECORANA ordination on P(:-ORD 2.0 (hijA1 Software Design. 1995). Preliininav data deso-ipiion F-ield ohsen ations suggested that plants represent four species D.j~clzszz, D. prnetemica, I). znmnakl and D.purpurella. An attempt was made to classify each plmt pa,/ hot front it\ recorded measurements, by converting each step in published taYonomic kc) \ into a formal logical algorithm which was applied to computerised data. For thk purpose. the labelluni Lvas defined as reflexed if the angle of fold e~ccedcd43 deLgrees.Three ke\s Mere used: Rose (1981), Clapham, ‘lutin & Moore / lW9) nnd Stace (1 997). Curiousl), feucr than half the sampled non-hybrid plants fittrd n whole sequence of key steps perfectly (Table 2), a proportion that would diiiost ccxrtainl) have been reduced further if additional key characteristics had been A\ ailable for analysis. A reasonable number of plants keyed down accurately to D. jurhtu or U . incarnnta. but ver) fekv matched D.praetermirra or D.purpurella. The commonest problem Mas for the stem to be recorded as solid for plants which would otherwise fit accuratcl) to the (supposedly hollow stemmed) spccies, reflecting thc dificult\ of assessing stem hollom ness non-destructively. Consequently, for the nna1)ses below plants \\..ill be identified as the species recorded in the field, even thouqh this introducrs an element of subjectivity. The distribution of putative qxcics between sites is given in Table 3. (The low numbers recorded from site 2 reflect poor establishment after translocation, due to drought and rabbit damage.) 4 summary of the univariate data for each named taxon is +en in Table 4. Clear differences between species are to be expected, 4ncc plants were allocated 10 species on the basis of predefined characteristics. (.llthough not shown, F values were calculated to test between-species differencrs for the data in Table 1 and were significant in all cases; these should be regarded ‘11 ‘I p o d hoc description rather than a true significance tcst.) MORPHOhIEI'RICS OF DACTYLORHI44 39 I TABLE 3. Brcakdown of species numbers by sites 1 2 3 4 5 z Uncertain D. fucluii D.praeteimicra 1). incamata D.puipurella 30 0 7 3 0 12 11 19 4 10 46 37 23 22 15 15 c 7 60 143 Site 11 1 5 0 0 0 1 0 3 7 0 57 8 11 4 0 0 TABLE 4. Summary results for univariate descriptive statistics, given as number of plants or as meanf SE. All lengths are in mm Species Total examined Spotted leaves Hollow stem Lip red Lip purple Lip white Lip dots only Height Number of flowers 1,eaf Icngth Lip length Folding angle Spur length Notch depth Uract/flower ratio Widest point ratio D. incarnata D. praetennissu 22 0 15 17 2 3 0 23 0 3 0 23 0 14 D. f u c h z 37 35 1 0 22 1.i 0 D. purpurella Hybrids 15 46 3 16 0 0 0 15 0 0 2 43 3 10 192.00k 18.00 400.001 18.00 339.00 f25.00 166.001 12.00 312.00 k 20.00 19.30 k 2.60 36.00k 3.40 36.705 3.10 16.30.t 2.10 29.005 2.60 81.00f 6.00 134.001 5.00 98.00f 4.80 64.50f 4.80 100.20f 3.90 7.10k 0.20 6.60.t 0.20 6.00k 0.20 6.20f 0.10 5.30f 0.20 64.00 f 2.00 41.00k 3.00 29.00k 3.00 35.00f 3.00 57.00.t 3.00 5.70.t 0.20 7.20.t 0.20 6.60.t 0.10 5.2Of 0.20 6.80k 0.30 0.70f 0.10 1.90.t 0.10 0.30k 0.10 0.801 0.10 0.50+ 0.04 0.73f 0.03 0.88.t 0.03 0.85f 0.03 1.03.t 0.05 0.975 0.04 0.50 0.01 0.51 .t 0.02 0.40k 0.02 0.41 k 0.01 0.30f 0.02 + The most distinctive taxon was D. incarnata; plants were almost invariably diminutive plants with a hollow stem, hooded, unspotted leaves widest near to the stem, and reflexed lips (most were D. incarnata subsp. coccinea (Pugsley) Sob, the subspecies with brick-red flowers (Bateman & Denholm, 1985; Stace, 1997)).The remaining species were more variable. Dacplorhiza purpurella was recorded only on the Leblanc waste site, consisting of relatively small plants with deep purple flowers and a barely notched or entire lip. None of these plants appeared to have a hollow stem, despite this characteristic being used in all the keys previously cited. Dacplorhiza pruetermissa was more variable; plants had unspotted leaves, a shallow notch and a purple lip that was often decorated with dots only (no lines or dashes). Again, many plants that otherwise matched this species accurately were recorded as apparently possessing a solid stem, despite a hollow stem being expected. Dacplorhiza fuchsii exhibited the greatest range of height, and was characterized by spotted leaves and relatively flat, deeply divided lips (often with a white ground colour). The uncertaidhybrid plants included almost the entire spectrum of variation recorded in the putative parents, except that none had the pure red flowers that characterize D. incarnata subsp. coccinea. Most had purple flowers whose lip was embellished with lines and dots, though two individuals on the Leblanc site were r - 1 WI.E 3. S u m m a n results of chi-square contingency table analyses of morphological characteristics of hyhrid orchids 1h 1-1 Sr * - Sr * RK I .K st St * - HI Hd RK I .R * + Ht * +- + Nt Hd * * + St * + .\ld,rt.viations: ' - '. significant iP<O.Oj; negative association betwrrn characteristics: ' ', significant (P<O.O5) positilc association between chara<-trristirs;* indic-ates cells o n the leading diagonal of the matrix. BK. bract ratio; Hd. hooded lcaf tip; Ht. height; I-.\, lip angle: LR. leaf ratio: Nt. Notch (sinus) depth; Sr, spur length, St, spottcd IC'l\.C~\. hyperchromic (Ettlinger, 1991) with a solid central mass of colour on the lip (one with an orange centre on a ground colour of purple, the other a purple centre on white). IVithin-plant variation in floral dimensions is not mentioned in standard works on this genus, but was evident in the populations studied. Of the seven continuous floral variables, two (labellum length and bract length) differed significantly (lY0.05) between flowers. The tendency was for the lowermost flower to have a significantly longer labellum and bract than the next flower up. Univariate evidence of clustering within the hybrid plants was sought by applying the Corner test of association to each independent pair of variables in turn (this in\.olved converting each variable into a new bistate variable, with 1 =above median and O=below median, entering these figures into contingency tables, and then applying the chi-squared test). The results are summarized in Table 5, showing seven si,;nificant results out of 36 tests (itself R 0 . 0 1 from a Poisson distribution \kith E= 36/20 = 1.8). The associations which were found to be significant are generally consistent with the morphometrics of the parent species. The observations that taller plants tended to have a deep labellar sinus and poorly hooded leaves, while spotted plants had a deeper sinus, suggest influences of D.fuchsii, whereas the tendency for shorter plants to have leaves widest closer to the stem suggests the parentage of D. incarnata. However, these patterns are weak, and it would be easy to dismiss the biometric properties within the hybrid swarms as wholly random. The Kruskal-Wallis test was applied to examine whether morphometric differences existed between plants on the Leblanc site and PFA sites. These analyses excluded hybrid plants (which must be assumed to be a heterogeneous grouping). The results are summarized in Table 6, showing a trend for plants from the Leblanc site to be $mailer than those growing in PFA. Interestingly, D.jkhsii and D.praetemisra both tended to have less reflexed lips at the Leblanc site. Alultivariate anabses The dataset was ordinated by Principal Components Analysis (PCA), showing that the first two axes (accounting for 28.7% and 19.8% of the total variance MORPHOhIETRICS OF DACTYI.ORHI<A 393 Significant differences between plants from PFA and Lcblanc sites, assessed by the Kruskal-Wallis test. Hybrids were excluded because they were assumed a priori to be a heterogeneous group, and Dacplorhiza purpurella was excluded as it did not occur on PFA TABLE 6. ~ D. incamata D. jkchaz Height Number of flowers Leaf length Lip folding angle Bract/flower ratio D. praetenzrsa *- ** *** - *- *- *- *+ Abbreviations: * K0.05; ** R0.01; '-', Lehlanc value i P F A value; significant results are left blank. '+', Leblanc value >PFA value. Non- TABLE 7 . Factor loadings from multivariate analyses. Abbreviations used: % V eigenvalues, expressed as variance except for DCA where the raw eigenvalues are given; BR: length of bract; F L distance from stem to tip of flower; HD: presence of a hooded leaf tip; HT: height; HY: lip hyperchromic; L1: leaf length; L2: length to widest point of leaf; LA: lipfold angle; LD: labellar depth; LN: lines on lip; L P vertical length of labellum; L W labellar width; N F number of flowers; PP: purpleness of ground colour; RD: redness of ground colour; SI: depth of labellar sinus; S L spur length; SM:stem solid; S P presence/absence of leaf spot PCA %V: BR: FL HD: HT: HY: LI: L2: LA: LD: IN: LP LW: NF: RD: PP: SI: SL SM: SP: DCA CVA (PFA 1) Axis 1 28.7 Axis 2 19.8 Axis 1 0.04 Axis 2 0.01 Axis 1 69.0 0.66 0.50 -0.46 0.77 0.03 0.79 0.71 -0.56 -0.14 -0.09 0.63 0.74 0.73 -0.59 0.33 0.45 0.35 -0.32 0.36 0.55 0.62 0.69 0.25 -0.13 0.23 0.42 0.59 0.54 -0.39 -0.03 -0.44 0.03 -0.02 0.46 -0.68 0.31 0.13 -0.70 101 124 223 - 23 142 7 86 212 171 109 117 - 13 -83 585 98 - 93 127 31 1 - 298 - 38 0.47 0.04 0.08 0.13 - 37 51 94 - 135 - 47 - 29 134 80 365 - 67 - 198 - 142 32 -61 - 129 -73 - 158 - 192 Axis 2 31.0 -0.32 -0.25 -0.15 -0.02 ~~ - I .OO -0.81 0.57 0.40 1.78 1.44 -0.97 - 1.01 0.10 0.70 0.12 0.03 0.53 0.63 -0.90 -0.06 0.61 0.43 0.49 -0.31 0.53 0.02 0.05 -0.10 0.87 0.49 -0.92 0.32 CVA (PFA 2) CVA (Leblanc) Axis1 62.8 Axis 2 37.2 Axis 1 58.5 Axis 2 25.4 0.45 0.82 0.27 0.67 -0.12 0.63 -0.01 0.34 -0.17 0.36 -0.44 0.80 0.26 - 0.07 0.37 -0.06 ~ ~ 0.28 -1.16 0.78 - 0.09 2.40 0.03 0.18 - 1.47 0.41 - 0.60 0.19 -0.10 0.98 0.00 -0.89 -0.74 0.77 0.0 1 -0.14 0.33 -0.24 - 0.32 -0.05 0.06 0.16 0.00 ._ ~ ~~ -0.87 0.78 4.07 -3.20 0.24 0.08 I .60 -0.57 1.05 -0.37 -0.25 -0.36 -0.17 ~ - 1.07 0.34 - 1.99 1.41 1.31 0.18 -0.80 -0.17 -0.13 0.44 0.87 -0.56 -0.02 ~ respectively) describe a roughly triangular layout of points. The factor loadings are given in Table 7. The first axis reflects size, as is normal for biometric studies (e.g. Sneath & Sokal, 1973). This ordination is shown in Figure 1, overlain with three different sets of symbols (Fig. 1A-C).Figure 1A shows the field-recorded species identities. Dacglorhiza incarnata forms one cluster at the negative end of the first axis (indicating diminutive stature), whereas D. jkhsii and D. praetermissa occupy the positive (taller) end of the first axis. The second axis separates D.fuchsii from D. -2 -1 0 1 First principal components axis y 2. t 2 1- I: 2 2 E5 L 3 Species ‘ - D purpurella 4 2 D praetermzssa 0 D. fuchszz D zrzcarnata t ?-hybrid .* 0- -1- a 3 c -2- $ e 1 -3 l - -3 I -2 I -1 0 1 First principal components axis I 2 Characteristics Lips without lines 4 Red flowers + None of the above 0 Spotted leaves u -3 -2 -1 0 1 First principal components axis 2 3 Hybrid locality Figure 1. Thc first t\vo ascs of a Principal Components Analysis applied to thr entire dataset. This sainr ,vaph is presented with thrcr diffcreiit labelling systcms: Labels indicate (A) provisional species identities; (Bj presence/absrnce of tasoiiomically si<gnificantcharacteristics; (C) the substratr identity of hyhrid plants. 395 6+ + t @ v lot0 I"- l 0 h 20 I 40 I 60 80 First DCA axis v D.purpurella 4 D. incarnata Species D. praetermissa 0D. fuchsii + ?-hybrid Figure 2. The first two axes of a Detre'nded Correspondence Analysis ordination of the entire dataset, with provisional species identities indicated. praetermissa; D. fuchsii clusters at the negative end of the second axis, whereas D. praetermissa occurs at the positive end. Figure 1B shows the presence of three taxonomically useful characteristics (spotting on the leaves, a lack of lines on the lip, and a red ground colour). This shows that, as a first approximation, leaf spotting corresponds to D.fuchsii, unlined lip to D. praetermissa and red flowers to D.incarnata subsp. coccinea. Leaf spotting was the weakest of these taxonomic indicators, with spotted-leaved plants occurring throughout most of the ordination space. Figure 1C shows the same diagram overlain with the distribution of putative hybrids. One exceptional individual from the Leblanc site plots clearly with D. &chsii but was recorded as a hybrid (shown as the black dot in the lower right-hand corner). This plant was identical to D.fuchsii in all respects except the flower pattern which was hyperchromic. The PFA hybrids show a weak tendency to occur in two clusters (upper left of the figure between D. praetermissa and D. incarnata, and central right, between D. praetermissa and D. &chsii), but the Leblanc plants exhibit no apparent pattern. The data were also ordinated by Detrended Correspondence Analysis. Factor loadings are given in Table 7. The first two DCA axes (Fig. 2) show a similar pattern to PCA. Again, the data points approximate to a triangle, with the positive end of the first axis corresponding to D. incarnata while the negative end contains D. fuchsii and D. praetermissa. Unlike PCA, there is no clear separation of D. fuchsii and D. praetermissa on the second axis. Parnell & Waldren (1996) warned that DCA is unstable when applied to bistate variables (such as leaf spotting, which was scored as 0 or 1 in this study), and suggested re-coding such variables to 0 or n where n is the total number of variables. In this case re-coding was found to have no effect on the ordination. Data were subjected to Cluster Analysis using Ward's method, nearest neighbour, furthest neighbour and UPGMA (Unweighted Pair-Groups Method, using arithmetic P. J. A. SHAII' + o+ 9 + $+*+ , ++++ + First discriminant function 4 I - 8 D. incarnata I I - 6 4 D. praetermissa I 10 - 2 0 U 2 4 I 6 8 First discriminant function Figure 3. Discriminant Function Analysis plot of biometric data for species plus hybrids from PFA sittx This graph is shown twice with two different labelling systems. Labels indicate (A) provisional species identities; (B) presence/absence of leaf spotting. Note that the 'D. fuduii' cluster is defined solel!- hy leafspotting. averages) algorithms applied to a matrix of Euclidean distances. The results are not displaved, since the dendrograms showed poor resolution ( judged by the field-based species identifications) and differed Miidely in their structure. Additionally, there are practical difficulties in presenting legible dendrograms containing over 100 individuals (Gauch, 1982). Discriminant Function Analysis was applied using plants identified in the field to named species as targets. This analysis explicitly assumes that the target individuals ha1.e been correctly identified (Kent & Coker, 1992), and uses their hiometric properties to allocate any remaining unclassified ('hybrid') individuals into the prespecified target groups. Consequently, this technique is less natural and more prone MORPHOMETRICS OF DAC'TYLORHIZA 397 61 + ++ + +@ 00 o+ 0 aoB0'0 8 0 I " - 6 + 4 + I 2 4 4 I I 2 4 First discriminant function - + + 1 4 & ++ ++ ++ ++++ 0 I 6 8 Species D. incarnata 0 D. fuchsii 0 D. praetermissa + ?-hybrid 4 Fi<gure4. The Discriminant Function Analysis shown in Figure 3, repeated after excluding leaf spotting from the dataset. Labels indicate species identities. to observer error than the previous analyses, but is a logical approach to the classification of the uncertaidhybrid plants. Separate analyses were performed for the PFA sites and the Leblanc site, since they contained different numbers of species (three and four respectively). Factor loadings are given in Table 7. The first discriminant analysis for the PFA sites incorporated all biometric variables except hyperchromism (which was excluded as it was only recorded on 'hybrid' individuals), and produced two distinct clusters, one corresponding to D. j k h s i i and the other containing a continuum from D. incarnata to D. praetemissa (Fig. 3A). However, this result proved to be highly misleading when the diagram was reinterpreted in terms of the presence/absence of leaf spots (Fig. 3B). The D.fuchsii cluster was uniquely defined by the presence of leaf spotting, and the other cluster by its absence. This result seems overly simplistic, and contradicts the wide spread of leaf-marked individuals within the PCA ordination (Fig. 1B). It arises as an artefact of the discriminant analysis: almost all definite D.fuchsli had spotted leaves, whereas none were recorded on other species on PFA, so spotted leaves were disproportionately represented on the first discriminant axis. The discriminant functions for the PFA sites were re-calculated after exclusion of leaf spots, producing a very different ordination (Fig. 4). This shows a trianplar pattern, with each corner corresponding to a named species and part of the intervening space being occupied by hybrid plants. The hybrids occupied two distinct regions of the ordination space: midway between D. praetemzissa and D. incarnata, and intermediate between D. praetermissa and D. fuchsii. There were no points in the space between B. fuchsii and D. incarnata suggesting that little if any hybridization occurred between these two species. Discriminant function analysis was applied separately to data from the Leblanc site, again with leaf spotting excluded (Fig. 5). This generated one well-separated cluster corresponding to D. incarnata; all other plants occupied an elongated cluster L V 4 4 * 444 r + + a *+ a- -12 - 10 , 20 Species I ~ 1 0 10 First discriminant function 3 D.purpurella 4 D Lncarnata D.praetermissa 0 D fuchsri f ?-hybrid Figtirc 5. Ilisc riiiiiiiaiit Function .\rial) sit plot of orchids from the Leblanc site (after ruclusion of leaf spc~ttiiigj.slio\\ ing yx%cic\idrntitirs. running from D. praeterinzs~uto D. purpuwlla. The few plants of D.praefermwa were kvell separated from the remainder, but there was overlap between D.3ch~ii,D. piqburdu and the hybrids. The abo\,e analyses produce a concise description of variation within a group of plants prone to hybridization. -4s such, it should be cxpected that there will be no unique correct answer, and the criteria for an acceptable set of results must include ~ k g a n c eand user satisfaction (Goodall, 1978). The ohsewation that univariatc analyses found ‘textbook’ differences between hpccics is trivial, since the field species identifications were based on cxpected combinations of characwrs. Of more interest are thc significant differences between planti from PF.4 and Leblanc sites (Tahlc 6), which show plants from the Leblaric sitc to be generally smaller and \\ith feuer flowers. This probably reflects the drier ndture of the Leblanc site (20 I n above local nater level, compared with PFA lagoons at. or just above. the Lvater table). Ilarploihzza furhsii and D. pruetermwu also had flatter floltcrs at the Leblanc site. although the cause of this is unclear. hlulti\.ariate techniques are confimed as powerful tools for obtaining an objective o\en-iew of morphometric datasets, and for examining the extent to which preconceived species definitions are 1,alid. PCA is the most mathematically natural ordination technique, so that the PCA ordination diagram can be used as an ohjccti\ e summar) description of the results. This diagram (Fig. 1A-C) approximates t o a triangle, with the corners corresponding to D.irzcarnata, D.praeternzzssa and D. j k h > ~ zLluc!ylorhi& . pzirpurella occupies a cluster midway between D. praetemxtu and MORPHOMETRTCS OI' DAC'T1LUKHIGl 399 D. incarnata, and hybrids are scattered unevenly throughout the ordination space. This result is an unbiased description of patterns within the data and (unlike the univariate analyses) confirms that accepted species descriptions correspond to extremes of variation within the Dacplorhiza colonies. It is possible to use positions within this ordination diagram to estimate the parentage of hybrid individuals, although this approach requires caution (Wilson, 1992). For PFA sites the hybrid plants appear to occur in two loose clusters: one lies between D. jkchsii and D. praetermissa, the other between D. praetermissa and D. incarnata. This pattern suggests that the hybrids recorded were mainly D. praetermissa x D. fuchsii (0.x grandis (Druce) P.F. Hunt) and D. praetemissa x D. incarnata (0.x wintoni (A. Camus) P.F. Hunt). No clustering of hybrid plants was evident on the Leblanc site. On both sites plants intermediate between D.fuchsii and D. incarnata were conspicuously lacking, in agreement with the observation by Hedrtn (1996a) that these two species appear to be genetically isolated. The results of DCA ordination generally mirrored those of PCA, but the separation was poorer than PCA, with only D. incarnata showing cohesion. The failure of cluster analyses to produce a coherent or consistent classification of the data is disappointing but not surprising. Cluster analysis is notorious for producing information of dubious quality (Cormack, 1971) and since it seeks clear divisions within dataspace it is ill-suited to partitioning data such as the results presented here which approximate to a smooth multivariate continuum (Gauch, 1982). Discriminant Function Analysis is a logical approach to the problem of classifying individuals from a mixed population, but its results here highlighted the importance of checking that the results of a multivariate analysis are consistent with user experience. When this analysis was applied to orchids from PFA sites it effectively generated an unbreakable rule that any plant with leaf spots was D.$chsii (Fig. 3B). This simplistic result did not match field notes, nor was it compatible with the PCA ordination. Excluding leaf spotting from the discriminant functions gave a result that coincided more closely with PCA, namely that hybrids at PFA sites were intermediate between either D. fuchsii and D. praetermissa or D. praetermissa and D. incarnata. Discriminant Function Analysis, when applied to plants from the Leblanc site, separated out D. incarnata and D. praetermissa, but left D. fuchsii, D. purjwella and hybrids in a diffuse cluster. Thus, two lines of numerical analysis (PCA and Discriminant Function) concur in suggesting that introgression has produced a more even filling of multivariate space on the Leblanc site than on PFA substrata. This may reflect the presence of a fourth species on the Leblanc site, combined with the substantially greater age of the habitat (c. 100 years, compared with c. 30 years for the PFA sites). Both factors allowed greater frequency of hybridization on Leblanc. Anderson (1995) reported greater levels of morphological diversity among Dacplorhiza traunsteineri in older populations, although in her example population ages were estimated (from geological factors) in thousands of years rather than decades. It is unlikely that the taxonomic problems of this group can be solved by morphometric techniques alone. Future work should aim to integrate morphometric measurements with chromosomal, protein and DNA-based techniques, to explore whether the phenotypic variability of these plants is matched by an equivalent level of genetic variability. ?'he author is grateful to Judy Xdams (Lee Valley Park Authority), Ian Fenton (Sational Pouer Drax). M'es Halton (hloses Gate Country Park), Alison John iC;round\\ ork Erewash) and Dennis Vickers (London Wildlife Trust) for site access m d hclpfiil discussions, and to the Roehampton Institute London for the granting of \tud! leave. Erica hfcAlister and Vince Gardiner commented on drafts of this inan uwript . REFERENCES Adcock EM, Gorton E, Morries GP. 1983. A study of some Dactvlorhzza populations in Greater .\lanchewr. I4btsonia 14: 377 389. Andersson E. 1994. 0 1 1 t h identit\~ of orchid populations: a morphomctric study of the Dactylorhira trm/n\t~i~~ri c-oniplex in eastcrn Sweden. .\ordic .j%unial 0fBotary 14: 269--275. Andersson E. 1995. Age-rclatcd morphological differentiation among populations of Dacglarhiza traimfrinrii (Orchidacracj in eastern Sweden. "IordirJournal OJ'Bohg15: 127-1 37. Bateman RM, Denholm I. 1983. .\ reappraisal of British and Irish dactylorchids. 1. 'I'he tetraploid marsh-orchids. I t khortia 14: 347 376. Bateman RM, Denholm I. 1985. l: reappraisal of British and Irish dactylorchids. 2. Thc diploid marsh-orchids. Ilbtmin 15: 32 1 -35.5. Bateman RM,Denholm I. 1989a. X rcappraisal of British and Irish dactylorchids. 3. The spottedorchids. I f h n i a 17: 3 19 ,349. Bateman RM, Denholm I. 1989b. Alorphometric procedure, taxonomic objectivity and marshmatics. i4ht.sonia 17: 449-455. Box J. 1993. C:onser\ation or greening? Thc challenges of post industrial landscapes. Britirh M/ildlije 4: '?73--27',. Burrows C. 1995. .\xi investigation into the vegetation at Nob End Leblanc Wastc Site, Bolton. L~npiiblishrd11% thesis, Lancaster University. Clapham AR, Tutin TG, Moore DM. 1989. Flora oJthe British Isles, 3rd edition. Cambridge: CUP. Cormack RM. 1971. A revicw of classification.Journal ofthe Royal Stati.stical Society A134: 321-367. Digby PGN, Kempton RA. 1987. .\Iultii'ariate inethodr in erologv. London: Chapman and Hall. Ettlinger DMT. 1991. Tw.0 ne\v varieties of British Dartvlorhiea. bI6tsonia 1 8 307-309. Gauch HG. 1982. .\fultirariut~ ana!i'sk zn rummuuig, eroloQ. Cambridge: Cambridge University Press. Gemmell RP. 1977. T?rr roloniratiori of indutrial umteland. IoB Studies in Biology 80. London: InJtitute aj Hidogs. Gittins R. 1985. (;anonzm/ ana!nz.~.a.srn.zmc. uith applications in e r o l o ~ .Berlin: Springer-Vrrlag. Goodall DW. 1978. Xumcrical classification. In: JVhittaker RH, ed. (:lassiJicution ofplant communities. '1'11~Ha,pe: Junk. 247--286. Greenwood EF, Gemmell RP. 1978. Derelict industrial land as a habitat for rare plants in S L a na ,\'C: 59j and I S Lancs a 7 C 60). Llbtmniu 12: 33-40, Hedren M. 1996a. Genetic dflerentiation, polyploidization and hybridization in northern European I)ai.hlorhi;o iOrchidacearj: evidcnce from al1oz)mc markcrs. Plant Systematics and Eoohtiou 201: 3 1 55. Hedrn M. 1996b. The. allotetraploid nature ofDar~i~lorhi~apraetermissa confirmed. Wat.ronia 21: 1 13 1 18. fill MO, Gauch HG. 1980. Detrcndcd correspondence analysis: an improved ordination technique. I&J/atio 42: 4758. Jeffers J. 1996. ~lulri\~ariate analysis of a reference collection of elm leaves. Journal of'ilpplied Stutistiu 23: 37 I .513i. Kent M, Coker P. 1992. I$g&tion dtscriplion and anabAir, a practical approach. London: Wiley. Kovach WL. 1990. .\IfXP 2.U-.ZlultiztariutP Statirtiral Package. Anglesey, UK: Kovach Computing Srn-ices. Lee JA, Greenwood B. 1976. The colonisation by plants of calcareous wastes from the salt and alkali indiistn in Cheshirc. England. Bzological Consm~atzon10: 131--149. MORPHOMETRICS OF DACTYLORHIZA 40 1 Mathews MH. 1978a. The decline of the British Leblanc alkali industry in the 19th century-a spatial perspective. Cambria 5: 46-68. Mathews MH. 1978b. The geography of the British heavy chemical industry in the 19th century. 'Tijdschrift voor Econ. en Sac. Geograjie 69: 333-344. MjM software design 1995. Pc Ord: multivariate ana$ses ofenvironmental data, version 2.0. Oregon: MjM software design. Parnell J, Waldren S. 1996. Detrended correspondence analysis in the ordination of data for phenetics and cladistics. Taxon 4 5 7 1-84. Rose F. 1981. n2e wildjower kq. London: Penguin books. Shaw PJA. 1992. Successional changes in vegetation and soil development on unamended Hy ash (PFA) in southern England. Journal OfAPplied Ecology 29: 728-736. Shaw PJA. 1994. Orchid woods and floating islands-the ecology of fly ash. Britirh Wldlaji 5: 149-157. Shaw PJA. 1996. Role of seedbank substrates in the revegetation of fly ash and gypsum in the U.K. Restoration Ecology 4: 6 1-69. Shaw PJA. 1998. Conservation management of industrial wastes. Practical Ecology and Conservation 2: 13-18. Sneath PHA, Sokal RR. 1973. Numerical taxonomy. San Francisco: Freeman. Stace CA. 1997. Newjfora of the British Isles (2nd ed). Cambridge: CUP. Steel RDG, TorrieJH. 1982. Principles andprocedures of statistics-a biometrical approach. London: McGrawHill. Wilson P. 1992. O n inferring hybridity from morphological intermediacy. Taxon 41: 11-23.