First instalment in resolution of the Banksia spinulosa complex (Proteaceae): B. neoanglica, a new species supported by phenetic analysis, ecology and geography

Abstract Taxa in the Banksia spinulosa Sm. complex (Proteaceae) have populations with sympatric, parapatric and allopatric distributions and unclear or disputed boundaries. Our hypothesis is that under biological, phenetic and diagnosable species concepts that each of the currently named taxa within the Banksia spinulosa complex is a separate species. Based on specimens collected as part of this study, and data recorded from specimens in six Australian herbaria, complemented by phenetic analysis (semi–strong multidimensional scaling and UPGMA clustering) and a detailed morphological study, we investigated both morphological variation and geographic distribution in the Banksia spinulosa complex. All specimens used for this study are held at the N.C.W. Beadle Herbarium or the National Herbarium of New South Wales. In total 23 morphological characters (11 quantitative, five binary, and seven multistate characters) were analysed phenetically for 89 specimens. Ordination and cluster analysis resulted in individuals grouping strongly allowing recognition of distinct groups consistent with their recognition as separate species. Additional morphological analysis was completed on all specimens using leaf, floral, fruit and stem morphology, providing clear cut diagnosable groups and strong support for the recognition of Banksia spinulosa var. cunninghamii and Banksia spinulosa var. neoanglica as species.


Introduction
Th ese competing taxonomic treatments have created confusion, examples of which can be found in species lists for some National Parks in New South Wales (unpublished visitor brochures), which include B. spinulosa var. neoanglica and B. cunninghamii subsp. A as separate entities. Some herbaria also concurrently use two names for the same entity (see the Atlas of Living Australia). Current circumscriptions of the taxa within the B. spinulosa complex are based on intuitive assessment of observed morphological variation, rather than an explicit analysis of the morphological variation. Th iele and Ladiges (1996) conducted a cladistic analysis of the whole of Banksia using 92 qualitative characters and 14 morphometric characters in an attempt to clarify interspecifi c relationships and to provide a phylogenetic classifi cation. As that was a genus-wide study, limited work was conducted on or within individual species.
Th e aims of this study were (1) to test and set the taxonomic status and circumscription of B. cunninghamii subsp. A; and (2) to search for novel diagnostic characters that could be used to distinguish individual taxa within the B. spinulosa complex (sensu George 1988).

Study material
Although dried herbarium specimens were available for this study, it was considered necessary to collect fresh material to adequately investigate character homology though a detailed study of diff erent developmental stages. Existing herbarium specimens were defi cient in some developmental stages and often were not suitable for destructive sampling. We made collections from locations in New South Wales and Queensland encompassing the full geographic range of Banksia cunninghamii subsp. A. Vouchers have been lodged at NE and/or NSW (Table 1). Each site was visited twice; the fi rst time in February to observe the development of the rachis, the second time in May to observe the fl owering process. During both visits observations were made and vouchers prepared.

Observations and microscopy
Micromorphology was examined using Leica MZ8 and MZ9 stereomicroscopes fi tted with eyepiece graticules. Images were taken using a Wild M400 photomacroscope fi tted with a Nikon DS-5M-L1 Digital Sight Camera System. Exploratory scanning electron microscopy of styles was undertaken using air and silca gel-dried samples mounted on double-sided carbon tabs on aluminium stubs. Specimens were coated with gold in a Neocoater sputter coater and examined at 15 kV using a Neoscope JCM-5000 bench-top SEM.

Selection of characters
Th e character list was primarily constructed to include leaf, fl oral, stem and fruit morphology. Assessment of descriptions of the taxa in the B. spinulosa complex (George 1981(George , 1988Th iele and Ladiges 1996;Harden 2002) led to the selection of characters for the inclusion in the phenetic analysis. Additional characters were considered based on observed diff erences in the fi eld (Table 2). Wherever possible, quantitative characters were used to reduce subjectivity and to avoid artefacts resulting from the conversion of continuous variables into categorical ones. Qualitative character states were scored as either 1 or 2. Quantitative characters for each OTU were the mean of up to 10 measurements where possible. Colours, however, were treated as multistate characters to maximise accuracy and repeatability, which allowed for some natural variation, thus avoiding spurious over- Length of complete confl orescence including peduncle ± 1 mm 2 Width of lamina at widest point excluding teeth ± 1 mm 3 Length of lamina including mucro ±1 mm 4 Length from base of lamina to fi rst tooth excluding mucro ± 1 mm 5 Length of seed including wing ± 1 mm 6 Width of wing at widest point ± 1 mm 7 Length of seed excluding wing ± 1 mm 17* Number of fl oral pairs around circumference of confl orescence 9 Length of complete infructescence ± 1 mm 10 Circumference of complete infructescence ± 1 mm 12 Lamina interveinal thickness when dry ± 0.05 mm Binary characters 11 Lignotuber: 1 = absent 2 = present 20* Floral bract keel number: 1 = 1 2 = 2 21* Distal bract margins: 1= plain 2 = recurved 22* Bract apiculum: 1= absent 2 = present 23* Bract apiculum: 1 = not incurved 2= incurved Multistate characters 8 Lamina apex: 1 = tridentate, 2 = bidentate, 3 = unidentate 13 Colour of lamina adaxial surface when dry1 4 Colour of lamina abaxial surface when dry1 5 Colour of lamina adaxial surface prior to drying1 6 Colour of lamina abaxial surface prior to drying1 8* Style colour pre anthesis1 9* Style colour post anthesis~ = RHS colours, see Table 3. * = new characters; i.e. not previously used in studies of Banksia (cf. Th iele and Ladiges 1996).
precision (see below). Royal Horticultural Society (RHS) colours were used to compare adaxial and abaxial leaf surfaces prior to, and after drying, as well as styles before and after anthesis. Each RHS colour was allocated a number from 1-26 (Table 3).
All leaf measurements were taken from leaves in the middle of a branchlet, selected from the whorl of branchlets subtending a resting terminal bud or confl orescence; leaves were measured after drying. Confl orescence characters such as number of fl oral pairs were counted live on the plant. Infructescences were measured vertically with a steel ruler and the circumference was measured with a sewing tape measure.
Infructescences were placed on a gas burner for 1-3 min then left on brown paper for two days in a dry place. Seeds were extracted using a pair of forceps and measured under a stereomicroscope using a calibrated eyepiece graticule.

Dataset
A dataset (Appendix 1) was maintained in Microsoft Excel and exported to PATN v. 3 for Windows (Belbin and Collins 2006). Th e characters were range-standardised and a distance matrix calculated using the Gower distance metric (Wills et al. 2000). Th reedimensional ordination plots were generated from the distance matrix using semistrong hybrid multidimensional scaling (SSH MDS) with 100 random starts and 200 iterations to minimise stress. Flexible UPGMA (Beta-value = -0.1) phenograms, 3D ordination scatter plots, and correlation of characters with ordination pattern (PCC) were produced directly within PATN. Th e criteria for circumscribing distinct taxa were: (1) the OTUs representing the putative taxa formed discrete groups that did not overlap those of any other groups of OTUs in both cluster and ordination analysis and (2) the OTUs within these groups showed an amount of morphological heterogeneity similar to that of the other putative species in the B. spinulosa complex included in the analysis (Plunkett et al. 2009). In total 23 characters were used, 11 morphometric, fi ve binary, seven multistate qualitatively coded morphological characters (Table 2).

Conflorescences
Th e confl orescences of all taxa in the B. spinulosa species complex consist of an elongate woody rachis that has three types of bracts. Below the base of the rachis on the short peduncle are the involucral bracts. Th e second type of bract is the common bract each of which subtends a fl ower pair on the confl orescence axis. Th e third type of bract, a smaller fl oral bract, subtends each fl ower in a pair (Johnson and Briggs 1975;George 1981;Th iele and Ladiges 1996). In the early stage of confl orescence development, fl ower pairs start to develop along the rachis basipetally. Th e fl owers emerge from each side of the fl oral bracts and above and below each large common bract. Bracts and fl ower pairs are arranged in vertical columns on the rachis. Th is pattern is visually enhanced with the development of styles. Th e vertical striping pattern remains until the perianth and the styles have senesced or fallen from the rachis (George 1981;Th iele and Ladiges 1996;Collins et al. 2009).

Structure of the perianth (floral pairs)
Th e perianth segments or tepals in Banksia each consist of a limb and a claw (Th iele and Ladiges 1996). In Banksia and most other Proteaceae the perianth is made up of four tepals (Wrigley and Fagg 1989;Weston 2006).

Structure of the style
Th e confl orescences in the B. spinulosa complex have the appearance of being a particular colour, i.e. black, red, yellow orange, or purple. It is the styles that are most boldly coloured with red, black, green, yellow or purple pigment, not the limb and claw (George 1981;Collins et al. 2009). All styles in the B. spinulosa species complex are hooked and extend up to 3 mm past the limb and claw just prior to anthesis. Th e distal part of the style is modifi ed as a pollen presenter and the stigmatic cavity is located at the apex of the style. Th e style is released from the limb upon anthesis (Th iele and Ladiges 1996; Weston 2006). All styles in the B. spinulosa complex have similar surfaces. Scanning electron microscopy was performed on the style surfaces and no distinguishing features were found.

Leaf morphology
All taxa within the B. spinulosa complex have leaves that are scleromorphic in texture, discolourous, and linear in shape. Th e indumentum on the abaxial leaf surface is felted and the midvein is raised on the abaxial surface of all leaves in all taxa within the complex. Continuous variation was found in the colour of adaxial and abaxial leaf surfaces both within and between populations of all taxa within the B. spinulosa complex.

Lignotubers
Th e term lignotuber refers to a woody swelling which may take the form of an extensive subterranean lignotuber, basal lignotuber, or an above ground lignotuber (Mibus and Sedgley 2000). Th e development of a lignotuber is considered to have evolved repeatedly in diff erent lineages in response to increased fi re frequency (Whelan and York 1998).

Phenetic analysis
Ordination ( Figure 1) and clustering ( Figure 2) Figure 2). Even when we reran the analyses excluding all the binary characters (Characters 11, 20-23; ordination and phenogram not presented), the same fi ve groups of OTUs were obtained, which, along with the very low stress value (Figure 1) indicate that the results are robust. Twelve of the 23 characters, including quantitative, binary and multistate characters had correlated more than 70% with the ordination (  Th is cluster of OTUs could represent a distinct species, but we will investigate this question and the broader relationship between B. spinulosa var. collina and B. spinulosa var. spinulosa further before making any formal taxonomic changes to these taxa. Slight outliers in the B. spinulosa var. collina cluster represent some discontinuous morphological variation, which we also plan to investigate.

Taxonomic inference
Given the consistent clear cut groups in the ordination and cluster analysis across a broad geographic and morphological range of OTUs (Table 1), we propose the following taxonomic arrangement, which we use hereafter in this paper: recognising Banksia cunninghamii subsp. cunninghamii as B. cunninghamii sensu stricto; recognising B. spinulosa var. collina as B. collina sensu lato; recognising B. spinulosa var. spinulosa as B. spinulosa sensu stricto; formalising Banksia cunninghamii subsp. A at species rank under the name B. neoanglica . Although the OTUs of B. spinulosa from the Morisset and Bouddi populations could be considered to constitute a distinct species on the evidence we present here, we refrain from recognising these populations as a distinct taxon until we have more thoroughly tested the hypothesis that they are part of an extensive hybrid swarm and searched for any additional populations that might provide evidence for integradation between B. collina and B. spinulosa.

Growth forms within the Banksia spinulosa species complex
Banksia cunninghamii sensu stricto is a single-stemmed tree to 7 m tall, and is non-lignotuberous. Banksia spinulosa sensu stricto forms a multi-stemmed, rounded shrub to 3 m high. Th e lignotuber is subterranean. Banksia collina sensu lato is a multi-stemmed upright shrub to 3 m tall, with a subterranean lignotuber (Harden 2002;George 1981).
Banksia neoanglica has a variety of growth forms ranging from small rounded multistemmed shrubs to single-stemmed trees. Th e growth forms of B. neoanglica appear to be related to the degree of exposure of plants to fi re. At sites where there have been no fi res for more than 15 years, such as at Binna Burra, Lamington National Park, Queensland and some parts of Gibraltar Range National Park, New South Wales (Pers. Comm. Justin Kreis Ranger Glen Innes National Park), B. neoanglica is a single-stemmed tree and exhibits all the traits of an obligate seeder such as a greater infructescence load and spontaneous opening of the follicles. In the tree form, B. neoanglica has a slight swelling at the base of the trunk just below the soil or there are epicormic buds which often develop into branches, well above ground level, similar to those of some eucalypts (Burrows 2008). Th e multi-stemmed form has a substantial subterranean lignotuber and requires fi re to open follicles and has a greatly reduced infructescence load.

Individual adult morphological features
Styles: Th e structure of the confl orescence, including perianth and styles is similar for all taxa in the B. spinulosa complex. Size, shape and colour of the individual parts of the confl orescence, however, diff er considerably across the species. Style colour in the B. spinulosa complex varies depending on the proportions of chlorophyll (green), carotenoid (yellow to orange), anthoxanthin (yellow) and anthocyanin (red to purple) pigments that develop in them (Grotewold 2006). Th e style colour in B. neoanglica, B. spinulosa sensu stricto and B. cunninghamii sensu stricto usually grades from red to maroon to purple during confl orescence development, then the style becomes discolorous at anthesis, with the apex becoming dark purple to black. Th is is a consistent character within and between populations of three species in the Banksia spinulosa complex. Th e exception is B. collina sensu lato which has concolourous green styles both before and after anthesis. We found no black-styled B. collina sensu lato within the geographical range of this project.
Th e style apex in B. cunninghamii sensu stricto seems to have substantially more anthocyanin pigment than either B. spinulosa sensu stricto or B. neoanglica. In B. cunninghamii sensu stricto we observed that the style length is usually longer than either B. spinulosa sensu stricto or B. neoanglica and is a similar length to B. collina sensu lato. Th e black pigmentation of the styles of B. cunninghamii sensu stricto starts to develop one third of the way along the style above the ovary. In B. spinulosa sensu stricto and B. neoanglica the dark pigmentation in the style develops one half to two thirds of its length above the ovary. In all populations in the B. spinulosa complex with the exception of B. collina sensu lato we observed what appeared to be yellowstyled confl orescences. Upon closer inspection they are green styled and appear to have less chlorophyll in both the styles and leaves than is found in B. collina sensu lato which is also green-styled. Green styled variants are found in less than 2% of any one population except in B. collina. Polymorphism is a common trait in Proteaceae where, for example, 40% of all species of Protea exhibit variation in the bract, style and perianth colour (Carlson et al. 2010). It is often unclear whether these variants are transient mutant individuals or this feature is a persistent polymorphism (Carlson et al. 2010). In the case of the Banksia spinulosa complex, however, the variants comprise less than 2% of a population and were not found in every population; therefore it is unlikely to be persistent polymorphism.
Perianth colour: Th e colours of the perianth in the B. spinulosa complex vary according to their developmental stage and their exposure to sunlight. Th e perianth colours can vary within and between populations in all four of the species in the B. spinulosa complex. Th e factor that seems to have the most infl uence on the perianth colour in the early stages of development is exposure to sun, often mediated by the position of an confl orescence on the outside or inside branches of the plant or by shading from other plants. In B. spinulosa sensu stricto, B. collina and B. neoanglica, the confl orescences that are exposed to full sun tend to have orange or yellow perianths. Th ose that are exposed to a limited amount of sun tend to be green. Th e perianth colour of B. cunninghamii sensu stricto is diagnostic for the species. At maturity the perianth always has a distinct pink hue and this colouring continues through to anthesis. Th e pink hue does not vary between and within populations of B. cunninghamii sensu stricto, nor does exposure to full sun or full shade eff ect the colour of the perianth at maturity.
Common bracts: Common bracts have been mentioned in previous studies (Johnson and Briggs 1975;Th iele and Ladiges 1996;George 1981) but the bract surfaces had not been mentioned before this study or used to draw taxonomic conclusions. Close examination, especially at early stages of development, of the abaxial surface of the common (or fl ower pair) bracts found them to have diff erences in shape, texture, colour, and surface ( Figure 3A-D) which covary in line with the entities recognised here (Figures 1-2) within the B. spinulosa complex. We will characterise these diff erences for use in future expanded phenetic analysis and description of taxa in Banksia. Floral bracts were not examined in detail in this study.
Involucral bracts: Involucral bracts appear to be taxonomically informative at the species level in the study group. Th e involucral bracts of B. cunninghamii sensu stricto are caudate with an abaxial 'spine' ( Figure 4A). Th e involucral bracts in B. spinulosa sensu stricto ( Figure 4B) are longer and more scleromorphic, with little or no hair and no external spine. In Banksia neoanglica ( Figure 4C) these bracts are more hirsute without an external spine and in B. collina sensu lato ( Figure 4D) the involucral bracts are shorter, have no external spine and limited hair. Th ere are diff erences in the distal and proximal portions of the involucral bracts ( Figure 4A-D) in each species that warrant further examination.

Taxonomic conclusions
Th e diversity of species concepts in the biological literature is an asset, not a liability when considering the Banksia spinulosa complex and are an integral part of biological theory. We have taken into account the co-varying morphological discontinuities, the phenetic species concept, geographical and ecological isolation and the biological species concept of reproductive isolation. Th e use of diff ering concepts has been useful in suggesting multiple lines of evidence for testing taxonomic boundaries in the Banksia spinulosa complex (cf. de Queiroz 2007). Clear taxonomic groups were obtained based on the results of the morphometric analyses and corroborated by new characters (cf. Th iele and Ladiges 1996) such as the abaxial surface of the common bract (Figure 3), the number of fl oral pairs around the circumference of the confl orescence and obvious diff erences in the involucral bracts ( Figure 4A-D). Additionally, given the ecological isolation, reproductive isolation and morphometric diff erentiation of at least three of the taxa in the B. spinulosa complex, there is a compelling case to recognise Banksia spinulosa sensu stricto, Banksia cunninghamii sensu stricto and Banksia neoanglica as separate species (Table 5). Banksia collina sensu lato is considered heterogeneous and in need of further study, but is not readily confused with B. neoanglica. Similarly B. spinulosa from the Morisset, Bouddi and Calga requires further study but is distinct from B. neoanglica.
Th e geographic distribution of Banksia neoanglica falls within the biogeographic region known as the "Macpherson-Macleay Overlap" of Burbidge (1960), which is a biogeographically distinctive and rich area (Crisp et al. 1999) with many species of plants and invertebrates endemic to the area.
Banksia cunninghamii sensu stricto and B. neoanglica have often been misidentifi ed because B. cunninghamii sensu stricto, on occasions, has a brown indumentum; B. neoanglica sometimes also exhibits browning on the abaxial leaf surface. Th is char- acter has been used in the past as an aid to distinguishing B. cunninghamii sensu stricto and the two other 'varieties' recognised at that time (George 1981;Harden 2002). Indeed, this attribute occurs in both B. neoanglica and B. cunninghamii sensu stricto. Drying of the specimens in both of these species can cause browning on the abaxial leaf surface. Th e browning of the abaxial leaf surface should not be used as taxonomic marker or an identifi cation tool.

Future directions
Disjunct populations in central and northern Queensland currently assigned to B. spinulosa var. spinulosa warrant inclusion in a more broadly framed analysis, as do the northern and southern populations of B. collina sensu lato and Victorian populations of Banksia cunninghamii sensu stricto. Th ere are also other populations of Banksia that clearly belong with the B. spinulosa group but are as yet unstudied. Further work is needed to enable suitable placement of these populations. Analysis using molecular data, together with expanded use of the novel characters presented here, would likely resolve these long-outstanding taxonomic issues.   (2002).
No specimens so labelled have been located in NSW, BRI, CANB or MEL herbaria after repeated searches. Alex George (pers. comm. 2010George (pers. comm. -2011 could fi nd no specimens in PERTH and he believes it likely that specimens were never distributed. Accordingly, we have nominated a neotype, collected from the same population as the type. Description. Shrubs with 2-8(-10) stems to 2.5 m from a lignotuber or trees to 7 m tall. Juvenile leaves: petiole 2-3.8 mm long; lamina narrowly obovate, 30-66 mm long, 5-11 mm wide, strongly dentate along full leaf margin, apex bidentate. Adult leaves: petiole 1.8-3.5 mm long; lamina linear, 43-75 mm long, 3-4.5 mm wide, occasionally toothed towards the usually unidentate, occasionally bidentate apex; adaxial surface glabrous, with colour after drying RHS greyed green group 195a-d; abaxial surface felted, colour after drying RHS greyed white group 156a-d. Involucral bracts subulate, thickened at base, 3-15mm long, grey-brown pubescent. Confl orescence 84-119 mm long, 70-85 mm diameter at anthesis; fl oral pairs 12-14(-16) around the circumference of the confl orescence axis. Common bract with a single thickened keel on the abaxial surface that extends from the apex of the bract down to the visible part of the base of the bract, distal margins slightly concave, apex rounded, indumentum villous, lower third of bract uniformly brown and upper two thirds uniformly green (fi g. 3A). Perianth 18-23 mm long, pubescent, yellow-orange at maturity but may be green, orange or yellow during developmental stages; limb c. 3.5 mm long; anthers c. 1 mm long. Style 25-38 mm long, apically hooked, colour grading from red to maroon to black just prior to anthesis. Infructescence 85-120 mm long, 35-45mm diam. Seed 15-19 mm long, including wing. Figure 6.
Dis tribution. Banksia neoanglica occurs on the McPherson Range, just north of the Queensland-New South Wales border, Mt Warning and the eastern edge of the New England Tableland southwards to near Hanging Rock, New South Wales. Figure 7.
Ecology. Grows in sandy soil on granite and acid volcanics, rarely on basalt, in Eucalyptus open forest (Figure 6), woodland and heath at altitudes of 850-1480 m. Th e species is sympatric with Banksia integrifolia subsp. monticola throughout its range, with B. marginata sensu lato on the Gibraltar Range and with B. conferta in the Daves Creek area.
Th e growth forms that B. neoanglica assume appear to be dependent upon the exposure to fi re (Whelan and York 1998). In areas where there have been no fi res for more than 15 years, such as Lamington National Park, Queensland, and some parts of Gibraltar Range, New South Wales (pers. comm. Justin Kreis 25 May 2010), a singlestemmed habit is found. Here, the lignotuber is present as a stem thickening just above or just below the soil surface, and branchlets may sprout from epicormic buds up to 30 cm above the ground. Th is single-stemmed form of B. neoanglica behaves like an obligate seeder with a heavy infructescence load and follicles open spontaneously without