Evidence-based review of the taxonomic status of New Zealand's endemic seed plant genera

Abstract

The nature, value and relative importance of different lines of evidence for deciding the circumscription and rank of genera are examined and discussed. First, I argue that the circumscriptions of genera (and higher taxa) should always be monophyletic, preferably with strong support from independent evidence. Second, the rank of genus is decided on more subjective grounds, but several criteria are helpful: phylogeny should be recoverable from the classification hierarchy, genera should be distinctive, ranks should not be redundant and familiar names should be retained if other criteria are met. I then apply these criteria to an evaluation of the circumscription and rank of 59 New Zealand endemic genera of seed plants accepted in two recent checklists. Fifteen genera (25.4%) were rejected and 16 (27.1%) are considered equivocal. However, my lower estimate almost certainly underestimates endemism at genus rank because the circumscriptions of several genera not evaluated here, such as Pachycladon, could change in the future to become endemic. One new combination is made: Sonchus novae-zelandiae (Hook.f.) Garn.-Jones, although work in progress by others is likely to lead to several more.

Keywords:

• taxonomy
• genera
• circumscription
• rank
• New Zealand flora
• endemism
• monophyly

I am fully convinced that species are not immutable; but that those belonging to what are called the same genera are lineal descendants of some other and generally extinct species, in the same manner as the acknowledged varieties of any one species are the descendants of that species.

Darwin (1859, p. 6)

Introduction

In the 21 years since Chase et al. (1993) provided the first molecular phylogeny of a large sample of flowering plants, a major reorganisation of their classification at family rank and above has been undertaken. This has largely stabilised as the APG classification (APG III 2009). In New Zealand, most major herbaria and databases have followed this international research (e.g. Breitwieser et al. 2012). Now, as DNA sequencing and phylogenetic analysis have become commonplace, attention has turned to classification at generic rank. In New Zealand, the classification of genera, and especially those that are endemic, is likely to be a focus of molecular systematics research for at least the near future.

However, in spite of the wealth of data from DNA sequencing and the relative rigour of cladistic analyses and classifications, disagreement remains about particular classifications. Competing solutions to taxonomic questions point to an evident problem for taxonomy. Differing conclusions may arise when taxonomists apply differing criteria, when they emphasise different lines of evidence or when they are not explicit about the nature and application of their evidence. Objectivity is only possible when hypotheses can potentially be falsified. Taxonomic hypotheses can be falsified only by measuring how well they accord with known facts, whether these relate to relationships or to similarity. Practitioners should at least be explicit about their methods, results and conclusions (Goldacre 2008). Where appropriate evidence is not presented explicitly, the conclusions drawn should be regarded as unsupported and their acceptance withheld.

In spite of the difficulties, taxonomy has been progressive in rejecting subjectivity, assertion and authoritarianism in classifications (Hennig 1950, 1966; Sokal & Sneath 1963; Wiley 1975, 1981; Hamilton & Wheeler 2008), but this progress has been incremental and is still not complete. Many authors have proposed objective criteria for scientific classifications (e.g. Hennig 1950, 1966; Wiley 1981; Entwisle & Weston 2005; Hopper 2009; Vences et al. 2013).

Goals of classification

The classification of plant species into higher ranked taxa has largely been guided by two goals that are not necessarily complimentary: natural classifications and stable classifications.

Natural classifications

Many taxonomists have sought to produce classifications that are natural, that is, classifications that reflect something real from nature (reviewed by Judd et al. 2008). Following Darwin (1859), this has increasingly come to mean classifications that accord with evolutionary history. Natural classifications are hierarchical because of the evolutionary process, and classifications that reflect that hierarchy can be interpreted to add valuable meaning and predictions. Insofar as a classification system reflects evolutionary history, taxonomy provides a framework for the interpretation of comparative biology. Recently, rapid progress in molecular phylogenetics has led to many changes in classifications, so that classifications have arguably become more natural, but at the cost of stability, at least in the short term.

The approach to natural classifications has moved from intuitive methods, through phenetic analyses to phylogenetics. But even when phylogenies are considered the basis of natural taxonomies, there is disagreement among practitioners about whether paraphyletic groups are or are not acceptable (e.g. evolutionary vs. cladistic classifications).

Stable classifications

Both taxonomists and users of taxonomic information require plant names to be stable (e.g. Brickell et al. 2008). Changes in classification at the rank of genus directly affect the binomial, which can have legal and economic costs for users, in addition to the inconvenience for non-taxonomists needing to keep up with the taxonomic literature.

Conflict between naturalness and stability cannot be resolved by agreeing to abandon either. However, I believe that natural classifications based on sound evidence will increase stability in the longer term. This is borne out by the three (so far) iterations of the APG classification, where changes at the family rank in APG III (2009) are relatively minor. Thus, stability is served when classifications are firmly and explicitly based on evidence, and when there is general agreement on the kinds of evidence that are most informative.

Scope of this review

In this review, I consider evidence-based objective and subjective approaches to generic classification and apply these to a consideration of the current status of New Zealand's endemic seed plant genera. In part 1, I discuss the value of several lines of evidence as they relate to common criteria for genus-level classifications, focusing on their value to decisions about both circumscription and rank. In part 2, I apply the most valuable criteria to an evaluation of the evidence for acceptance of the 50–60 New Zealand seed plant genera that are currently recognised as endemic (de Lange & Rolfe 2010; Breitwieser et al. 2012).

Although classification criteria can be applied at all ranks, my focus is on the rank of genus, for three reasons. First, the genus is the highest rank that many individual taxonomists are able to research because taxa at family level and above often require research that is world-wide in scope, involving international collaboration and often big budgets. Second, genera are especially important because the genus is the only rank above species that affects the binomial. Thus generic names are also a shorthand way of indicating low-level relationships and, conversely, changes in generic circumscription affect nomenclature. Third, genus is often the rank at which ecological and evolutionary comparisons are conducted (e.g. Webb et al. 1999; McGlone et al. 2001; Lee et al. 2011), so that criteria for taxonomic decisions about genera can have major effects on research findings and on dependent policy (such as conservation priorities). Humphries & Linder (2009) argue that the genus may be the most important rank ‘to get right’.

In addition, although I briefly discuss the advantages of monophyletic classifications and mention the Phylocode, these topics are discussed in detail elsewhere.

Part 1. Principles of generic classification

Properties of taxa

In biological classifications, taxa have two fundamental properties: circumscription and rank. Taxonomists who circumscribe narrowly or who tend to raise taxa to higher ranks may be characterised as ‘splitters’, whereas those who circumscribe broadly or tend to reduce the ranks of groups may be characterised as ‘lumpers’ (de Queiroz & Gauthier 1994; Endersby 2009).

The principles of classification have been widely discussed throughout history (reviewed by Judd et al. 2008). Here, I refer mostly to three recent papers (Entwisle & Weston 2005; Hopper 2009; Vences et al. 2013) that set out criteria for the taxonomic recognition of genera and higher taxa, some of which are objective, but others of which are subjective. Entwisle & Weston (2005) reviewed criteria for taxonomic recognition with the aim of providing a single preferred classification of Australian plants through the Australian Virtual Herbarium (AVH). Following consultation and discussion within the Australian botanical community, they arrived at a set of recommended guidelines, of which the paramount one was that taxa should be monophyletic. Hopper (2009) considered these questions with special reference to recent ‘taxonomic turmoil’ in Australasian orchids. Vences et al. (2013) discussed criteria for Linnaean classifications in an attempt to limit changes within a consistent and cladistic framework; although their examples were mostly drawn from herpetology their conclusions are widely applicable.

These three papers are largely in agreement in concluding that all groups recognised should primarily be monophyletic, and then setting out a number of secondary, and more subjective, criteria for deciding which of many clades to name and rank. Here, I organise these thoughts under three headings: evidence for deciding (1) group circumscriptions, (2) formal taxonomic recognition of circumscribed groups and (3) ranks of recognised groups.

Criteria for deciding taxon circumscription

Species and higher ranks differ in the criteria used for their circumscription. Speciation is the point at which net-like relationships among individuals in a panmictic population, or panmictic populations in a metapopulation, give way to (mostly) branching relationships of species in a phylogeny. While genetic evidence about the circumscription of species falls largely in the domain of population genetics, evidence about higher taxa falls in the domain of phylogenetics.

Circumscription of genera and higher taxa concerns which species belong to a group and which are excluded. A group may be circumscribed by (1) listing the component individuals or subordinate taxa, (2) specifying some unique characteristic that members display (a character-based definition) or (3) defining relationships at its basal node or in relation to their most recent common ancestor (node- and stem-based definitions) (de Queiroz & Gauthier 1994).

Most modern botanists would agree that members of a taxon should be closely related to each other (Judd et al. 2008). However, there is disagreement about whether such within-group relationships should always be closer than any relationships to species outside the group (Brummitt 2002, 2006; Wiley 2009; Hörandl & Stuessy 2010; Schmidt-Lebuhn 2011).

Willi Hennig (1950, 1966) was the first to recognise that the only evidence for historical relationships between species is provided by synapomorphies, uniquely derived character states that are shared by related species. Then, having provided a method by which clades of historically related species could be discovered, Hennig proposed that only groups that are circumscribed to be congruent with clades (monophyletic taxa) should be recognised in taxonomic classifications. Hennig's view that only synapomorphies are informative in systematics (Hennig 1950, 1966; Wiley 1981) contrasts with the phenetic approach of using overall differences and similarities—the full extent of character-state differences between groups—which can be misleading about relationships (Wiley 1981).

Ever since Hennig (1950, 1966), the idea that taxonomic groups should be circumscribed so as to be monophyletic has been controversial, in that arguments have been raised against it, sometimes questioning the requirement for monophyly itself (Cronquist 1987; Brummitt 2002, 2006), and sometimes in reference to inconvenient nomenclatural changes that may result from its implementation (e.g. Brickell et al. 2008).

Practical importance of monophyly

Users of taxonomy are likely to assume that members of a named group, particularly a genus, are more closely related to each other than they are to members of another group. For example, the popular ecology software Phylomatic assumes that genera are monophyletic (Webb & Donoghue 2005). Thus, it is a desirable property of taxa that every member should be more closely related to every other member than any is to any non-member. This desirable and useful property is found only in monophyletic groups (clades).

Although alternative meanings of the term monophyly have been proposed by some (e.g. Ashlock 1971; Hörandl & Stuessy 2010), most systematists tend to follow Hennig (1950, 1966; Wiley 1981) who defined a monophyletic group as comprising an ancestral species and all of its descendants. Hennig (1966) also provided, more pragmatically, a clear statement of what monophyly means for a classification. In general terms, this can be restated: for any two monophyletic genera, A and B, every trio of species with at least one from each genus will demonstrate a three-taxon relationship of either ((A,A)B) or (A(B,B)). Such underlying relationships are historical, real and potentially discoverable facts.

Non-monophyletic taxa can seriously skew inferences drawn from classifications, because at least some of their relationships will be of the form ((A,B)A). Restricting taxonomic recognition to only monophyletic groups produces classifications that represent evolutionary history.

Using monophyly as a criterion for questions about circumscription has an additional advantage: monophyly is a hypothesis that can objectively be tested. Although the merits of different data sets (e.g. morphology, nuclear vs. chloroplast DNA, secondary metabolites) and the kinds of phylogenetic analyses (parsimony, likelihood, Bayesian) are debated fully elsewhere (e.g. Shaw et al. 2005; Hughes et al. 2006; Joly et al. 2009; Leaché & Rannala 2011; Straub et al. 2012), it is sufficient to note here that any data, considered in a phylogenetic context and capable of falsifying a phylogenetic hypothesis, are useful. Because of this objectivity and its explicit linking of systematics with evolution, many botanists now consider monophyly (sensu Hennig 1966) to be an essential property of higher taxa (Entwisle & Weston 2005; Hopper 2009; Schmidt-Lebuhn 2011).

Vences et al. (2013) argued additionally that monophyletic groups should have a stronger case to be recognised and named if they have clade stability, most likely in clades that have strong, independent, and congruent support from several lines of evidence. Some data sets (e.g. nuclear DNA vs. chloroplast DNA vs. morphology) can give conflicting signals about relationships (e.g. in Anemanthele [Poaceae] and Lophomyrtus [Myrtaceae], see below). Such conflict can be informative if evolution has been reticulate (e.g. in Poaceae subf. Danthonioideae, see Linder et al. 2010), or it may indicate that taxonomic conclusions should not be drawn until conflicts are resolved. In either case, uncritical a priori privileging of one data set over another is unwarranted.

However, because there is only one correct solution to the question of the history of life, it is theoretically possible to determine the phylogeny and test the monophyly of all taxa. In practice, it seems unlikely at present that we can confidently discover that solution exactly. Phylogenetic analysis of DNA sequence data is a theoretically sound approach that has also been shown empirically to be powerful in spite of difficulties associated with detecting reticulate evolution, potential lack of congruence among gene trees, the inference of species trees, incomplete lineage sorting, insufficient time for evolution of informative differences, and too much time so that informative differences are obliterated. Because many of these difficulties are especially relevant to studies of the New Zealand flora, some critics suggest the requirement for monophyly may be impractical (e.g. de Lange & Rolfe 2010) or dogmatic (de Lange 2011a). Nevertheless, in spite of many potential sources of error in tree-building, tree topologies—including reticulate ones—are testable hypotheses, so that errors in topology, branch lengths and support values can be discovered and corrected. Furthermore, improvements to gene and even genome sequencing and sampling, rapid sequencing, new analytical methods and computing are continually improving the power, speed and resolution of analyses (Soltis et al. 2011). Already there are many deep and shallow, species-rich and species-poor branches on the tree of life that we can recognise as monophyletic taxa with a high degree of confidence (e.g. angiosperms, Onagraceae, Fuchsia).

Non-monophyletic taxa

By recognising in formal classifications only those groups that are monophyletic, taxonomists avoid naming groups from which incorrect or misleading relationships could be inferred. In such non-monophyletic groups, some group members will be more closely related to non-members than they are to other (named) group members (Fig. 1). Non-monophyletic groupings are of two general kinds.

Figure 1 A simplified three-taxon statement about three shrubby New Zealand daisies, with their current generic classification (top) and their phylogenetic relationships according to Wagstaff et al. (2011, bottom). A, Olearia colensoi; B, O. solandri; C, Pachystegia insignis.

First, polyphyletic groups are those that occur when distantly related but similar organisms are classified together due to a misinterpretation of the origins of their similarity. Typically, character states initially assumed to be homologous may be found to appear on multiple lineages and to be the result of convergent or parallel evolution. The dismantling and reclassification of polyphyletic groups to restore monophyly is generally non-controversial. For example, Pseudopanax (Araliaceae) as circumscribed by Philipson (1965) was shown to be polyphyletic by Mitchell & Wagstaff (1997) leading to the segregation and reinstatement of Raukaua (Mitchell et al. 1997).

Second, paraphyletic groups occur when a distinctive subclade is excluded based on its differences and in spite of its relationships. The logic of rejecting paraphyletic groups has been widely and largely uncontroversially applied in recent New Zealand research (e.g. Heenan 1998c, 2002, 2007, 2012; Heenan et al. 2002, 2008; Wagstaff & Wege 2002; Garnock-Jones et al. 2007; de Lange 2012). In most examples, a subclade had been segregated at the same rank from within a more inclusive clade in order to recognise its distinctiveness, leaving the name of the original clade to apply only to a paraphyletic residue. Thus the distinctive Neopaxia australasica (Montiaceae) was segregated from Montia on the basis of its morphological distinctness (Nilsson 1966; McNeill 1975). Later this monotypic genus was greatly enlarged by the description of seven new species (Heenan 1999a), and when shown by O'Quinn & Hufford (2005) to be nested within Montia all species were returned to Montia (Heenan 2007).

Similarly, Lignocarpa and Scandia (Apiaceae) were segregated from Anisotome by Dawson (1961, 1967) on the basis of distinctive morphological differences. There is little doubt that these segregate genera are distinctive and monophyletic, but if they render Anisotome paraphyletic (Mitchell et al. 1998; Radford et al. 2001; Spalik et al. 2010; but see Nicolas & Plunkett 2009) their status should be reconsidered (see below).

In other examples, newly discovered species have been described in new genera, although some have been later found to belong to clades that already have names at genus rank. Thus, Garnock-Jones & Johnson (1987) proposed a new genus, Iti, for a newly discovered plant in Brassicaceae because they could find no synapomorphies that would place it in a known genus. Although they noted similarities to Cardamine, they excluded it from that genus because of its incumbent radicle and absence of rapidly coiling silique valves. Using explicitly phylogenetic methods and molecular data, Mitchell & Heenan (2000) and Heenan (2002) showed that Iti is nested within Cardamine (now as Cardamine lacustris) and has presumably secondarily lost the explosive fruit dehiscence (Vaughn et al. 2011) characteristic of that genus. Similarly, Carmichaelia (Fabaceae) has been enlarged by the inclusion of Chordospartium, Corallospartium and Notospartium (Heenan 1998a,c; Wagstaff et al. 1999) because excluding them rendered its circumscription paraphyletic. In both these examples, the influential morphological characters were related to functional changes to fruit dehiscence.

When a group is recognised to be paraphyletic, cladistic taxonomists can address it in classifications in two ways. Either the paraphyletic group can be enlarged to include the excluded subclade, or it can be dismembered into smaller monophyletic groups. This is a question of rank, rather than circumscription, and is addressed below.

Methodologically, the accidental recognition of groups that are paraphyletic is a potential consequence of phenetic analyses, that is, those analyses that estimate relationship based on an assessment of overall similarity. This is so regardless of whether phenetic analysis is carried out intuitively or by an algorithm. Pioneering pheneticists (e.g. Sokal & Sneath 1963) sought explicit and objective methods for classification. Many initially considered that quantifying overall similarity was an effective way to infer phylogenetic relationships, and this was generally thought achievable if sufficient morphological or genetic characters were analysed. Hennig (1966) and Wiley (1981) demonstrated the difficulty of inferring historical relationships through overall similarity, by showing that in phenetic classifications symplesiomorphies are as influential as synapomorphies, and they make likely the inadvertent recognition of groups that turn out to be paraphyletic.

Arguments for paraphyly

The proposition that taxa above species rank should always be monophyletic is widely but not universally accepted in the botanical community (for dissenting views see, e.g. Cronquist 1987; Brummitt 2002, 2006; Brickell et al. 2008; de Lange & Rolfe 2010; Hörandl & Stuessy 2010).

During the twentieth century, many, often small, new genera were segregated from large and variable genera, for example, in Bellis and Crepis (Asteraceae) (Humphries & Linder 2009). These were usually delimited on the basis of morphological distinctiveness, and with hindsight this justification can be seen to have been largely phenetic. However, the advent of phylogenetic analysis has shown that commonly such distinctive clades are nested within larger and also distinctive grades, such as Embergeria and Kirkianella within Sonchus (Asteraceae) (Kim et al. 2007). In such cases, the distinctiveness of the basal grade results in part from symplesiomorphy, relative to the apomorphies of the segregated genus. Many recent studies of New Zealand plants (e.g. Heenan 1998c, 2007; Chung 2007; Wanke et al. 2007) have rejected generic status for such distinctive subclades and merged them with the larger clades within which they have originated.

In spite of widespread acceptance of a requirement for taxa to be monophyletic, proponents of paraphyletic taxa in classifications (e.g. Cronquist 1987; Brummitt 2002, 2006; Hörandl & Stuessy 2010) argue that grades may be meaningful, even natural, because they represent stages in evolution that may be morphologically similar and occupy similar habitats or geographic areas. Schmidt-Lebuhn (2011) provided a detailed rebuttal of such arguments for taxonomic recognition of paraphyletic groups.

Further, if non-monophyletic groups are sometimes acceptable, are there any general and objective principles that could help us objectively to decide which ones to accept and name? Stuessy & König (2008) and Hörandl & Stuessy (2010) advocate patrocladistics, a system that uses phenetic distances and cladistic branching trees to derive classifications (see Wiley 2009 for rebuttal). It seems that some arbitrary amount of phenetic difference could establish a cut-off point, such that decisions might be supported by one data set but not by others, which could be a particular difficulty in fast-evolving island floras where morphological change appears to have proceeded faster than DNA sequence evolution (Winkworth et al. 2002a).

Criteria for deciding which clades to name

Once robustly supported clades have been discovered, an additional concern in a nomenclatural system with a limited number of available ranks is deciding which of them should be formally named. In any flow diagram or decision chart (e.g. Vences et al. 2013) this decision must follow circumscription of clades and precede rank assignment and naming.

Recognition

It is practical to expect named taxa to be distinctive, or at least diagnosable. This is particularly important to field biologists who might encounter organisms that they do not immediately recognise. At present, members of groups recognised only by distinctive DNA sequences cannot be identified in the field, although technology might enable this in future.

Naming only groups that are morphologically distinctive is subjective, because there are no explicit criteria for deciding which clades are distinctive enough to be named. Further, distinctiveness can be applied only after clades have been circumscribed, because it otherwise introduces the risk of recognising paraphyletic taxa (groups recognised by symplesiomorphies). For example, Norton & Molloy (2009) and de Lange & Rolfe (2010) argued that including southern segregates in a broad circumscription of Veronica (Plantaginaceae) (Garnock-Jones et al. 2007) does not draw attention to well-marked clades and does not recognise the distinctive characters that distinguish the Hebe group, and particularly Heliohebe, from a paraphyletic northern circumscription of Veronica. In this case, recognising well-marked clades as genera also dictates either the recognition at the same rank of clades that are not well-marked (e.g. Veronica clades I–IV of Albach & Chase 2001), or the recognition of a mostly well-marked grade (the standard 1929–2007 circumscription of Veronica). Further, if such phenetic criteria were applied strictly, it would likely indicate the separation of a number of plesiomorphic New Zealand and Australian (Oliver 1944; Albach & Briggs 2012) species from their nearest relatives. As an additional complication, however, the generic segregates of the Hebe complex are mostly neither monophyletic nor unambiguously well-marked (Garnock-Jones & Albach 2007); the exceptions are the sun hebe clade (Garnock-Jones 1993b, as Heliohebe) and the semi-whipcord hebe clade (Bayly & Kellow 2006, as Leonohebe).

Although it is desirable that recognised groups be distinctive, it does not necessarily make sense that distinctive groups should always be named, especially if their sister groups are not distinctive (Hopper 2009).

Biogeography and ecology

Vences et al. (2013) discussed the desirability of groupings that have a geographically coherent distribution or inhabit a similar environment. They regarded these criteria as ambiguous and difficult to apply. In addition, they should not over-ride a requirement for monophyly; for example, Mark (2012) and Norton & Molloy (2009) were influenced by their southern distribution in recognising Hebe and segregate genera as separate from a paraphyletic circumscription of Veronica.

International perspective

In New Zealand, high species endemism means that most taxonomic research at species rank can be conducted with little reference to the rest of the world, but independence becomes increasingly problematic at higher ranks, especially at family rank and above where no groups are endemic. At genus rank, New Zealand botanists should inform their decisions by considering plants from Australia, the Pacific Islands, Southeast Asia and further abroad. To this end, a DSIR–CSIRO colloquium in 1983 (Department of Scientific and Industrial Research 1987) led to increased international collaboration, particularly with Australia, and more multi-author research. Such collaboration and the improved availability of herbarium loans, international travel and electronic communication have led to taxonomists increasingly interpreting the New Zealand flora in relation to relatives overseas (e.g. Cross et al. 2002; Heenan et al. 2002; Garnock-Jones et al. 2007; Pelser et al. 2007; Knox et al. 2008; Linder et al. 2010; Smissen et al. 2011; Woo et al. 2011). It seems likely that studies conducted overseas will increasingly include New Zealand samples, and improved sampling generally will allow identification of the close international relatives of New Zealand plants (e.g. Barker et al. 2007a; Chung 2007; Kim et al. 2007; Wanke et al. 2007; Nazaire & Hufford 2012; Nicolas & Plunkett 2012).

Criteria for assigning ranks to recognised taxa

The rank of a taxon concerns where it is placed in a hierarchical system. Although the circumscription of monophyletic groups is consistent, objective and testable, assigning rank to a circumscribed group is for the most part subjective (Entwisle & Weston 2005; Hopper 2009). The subjectivity of ranks has led to proposals for unranked systems of nomenclature (brief outline and key references were presented by Laurin & Bryant 2009), which are outside the scope of this review.

Constraints of hierarchical classifications

First, the traditional codes of nomenclature, excluding the rankless Phylocode (de Queiroz & Gauthier 1994; Laurin & Bryant 2009) (e.g. the International Code of Nomenclature for Algae, Fungi, and Plants [ICN], McNeill et al. 2012), require ranks to be hierarchical; this places some objective constraints on their use. For instance, higher ranks cannot be less inclusive than included lower ones (e.g. families within a genus would not be allowed). That requirement rules out some classification options, but does little to guide the application of ranks and names to clades. Nevertheless, the genus, as a rank between species and family in the classification hierarchy, is constrained by the application of those higher and lower ranks if taxa are to be monophyletic. Thus, Notospartium (Fabaceae) could not be recognised as a monophyletic genus within the genus Carmichaelia, but could take a lower rank, for example, as a section, so long as all other sections were also monophyletic. Under the Phylocode, however, Notospartium could be a clade included within a monophyletic circumscription of Carmichaelia because neither clade would be assigned a rank.

Second, sister taxa ideally ought to be accorded the same rank (Hennig 1950, 1966) to recognise their common origin. This quickly becomes problematical if applied to a bifurcating tree because many more ranks are needed than are available if every node is named, so pragmatic solutions must be sought (Wiley 1981), such as assigning the same rank (e.g. genus) to each of the major subclades of a larger clade (e.g. a family) even when they do not all originate from nodes at the same level in the tree.

Rank redundancy

While adherence to the notion of naming every clade in a phylogenetic classification may be impractical, classifications ought to maximise information content such that a classification reflects phylogeny as closely as possible and never misrepresents it (Nelson 1973). In addition, redundant ranks should be avoided where possible. Ranks are redundant when two taxa exist that have identical circumscription but different ranks (e.g. Nothofagus s.l. and Nothofagaceae). The elimination of redundant ranks can improve the information content of a classification. Monospecific genera and families convey no relationship information, although they might sometimes be unavoidable (Entwisle & Weston 2005). For example, when two sister genera are monospecific and classified together with other genera in a larger family, like Manoao and Lagarostrobos within Podocarpaceae, there is no rank in the classification hierarchy that represents their relationship (Kelch 2002). The alternative, recognising these two sister species within one genus Lagarostrobos, is more informative, even though, as Molloy (1995) has shown, there are distinctive morphological differences between them.

Hymenanthera and Melicytus (Violaceae) were combined under the latter name by Green (1970) and Garnock-Jones (in Connor & Edgar 1987), following Beuzenberg (1961). Recent phylogenetic study (Mitchell et al. 2009) suggests there may be two well-supported sister clades within Melicytus that closely match the former circumscriptions of the two genera, although the position of M. lanceolatus is unclear. Mitchell et al. (2009) preferred to continue to recognise a broad circumscription of Melicytus. Nevertheless, further research should test the idea that two distinctive clades could be recognised at genus or subgenus rank, a solution that would add information content if monophyly of both could be demonstrated.

The family Nothofagaceae provides a third example. The order Fagales has two subordinate clades. One of these comprises seven families and is sister to the other, which includes only the family Nothofagaceae with one genus, Nothofagus (Li et al. 2004). While family and, potentially, suborder ranks convey information about its sister clade (Fagaceae, Betulaceae, Myricaceae, Ticodendraceae, Rhoipteleaceae, Juglandaceae and Casuarinaceae), the traditional circumscriptions of Nothofagus and Nothofagaceae are congruent and one rank is thus redundant. The recent subdivision of Nothofagus (Heenan & Smissen 2013) recognises well-defined clades (previously treated as subgenera within Nothofagus) at the rank of genus. New names that result from recognition of these former subgenera at genus rank instantly convey the information that silver beech (Lophozonia menziesii) is not the closest relative of the other New Zealand southern beeches (Fuscospora).

Nomenclatural stability

There are pragmatic conflicts in the argument for continuing to recognise distinctive groupings that are not clades. Although it would suit some users of classifications, for example, horticulturalists who might have real costs in re-labelling stock for sale (Brickell et al. 2008), legislators who wish to prohibit drug plants or protect plant variety rights, or botanists who worry about the effects of nomenclatural changes on the interpretation of old publications, acceptance of non-monophyletic groups can introduce major distortions into research or applications that depend on evolutionary relationship, such as in ecology and drug discovery.

For this reason, Entwisle & Weston (2005), Hopper (2009) and Vences et al. (2013) argued that although named taxa should always be monophyletic, ranks should be applied to clades in such a way that existing nomenclature is disturbed as little as possible, consistent with the above criteria.

Hybrids

Although Entwisle & Weston (2005) did not believe that the existence of intergeneric hybrids is evidence for or against generic segregation, where they do occur they are an indication of close genetic similarity or functional compatibility of genomes. Although a strict application of the biological species concept would suggest that ability to interbreed indicates two plants are conspecific, most taxonomists sensibly will accept low levels of hybridisation between species, particularly where hybrids have reduced fertility. Does hybridisation between plants in different genera provide us with any useful information regarding their evolutionary history and generic classification?

An intergeneric hybrid should not be viewed as a hybrid between two genera, but as an interspecific hybrid between two plants that are classified in different genera. The further back in time the last common ancestor of two plants occurred, the lower the likelihood that they can successfully interbreed. Because the ability to hybridise indicates close genetic similarity and likely indicates recent common ancestry, the existence of intergeneric hybrids is a strong suggestion that generic boundaries might be incorrectly delineated, although it should be borne in mind that the similarity might reflect shared plesiomorphic features of the genome. Morgan-Richards et al. (2009) suggested that some intergeneric hybrids in the New Zealand flora might involve species from recent species radiations.

de Lange & Rolfe (2010) took a different position, arguing that the low or zero fertility of some hybrids within the New Zealand Veronica complex (Veronica sect. Hebe–Plantaginaceae) was evidence that supported their generic level separation as Chionohebe, Hebe, Hebejeebie, Heliohebe, Leonohebe and Parahebe. Further, de Lange & Rolfe (2010) argued that the absence of hybrids between V. sect. Hebe and northern Veronica was evidence they should not be congeneric. In my opinion, such a generic concept is not biologically realistic. First, it would leave the defining characteristics of genera very close to those of (biological) species, because it implies that all congeneric species should be able to form fertile hybrids. Second, such a criterion would lead to a big reduction in the circumscription of most genera and many would be monotypic. Third, it is an asymmetric test, because while ability to hybridise indicates genetic similarity and compatibility and this implies some level of close (congeneric?) relationship, failure to hybridise does not falsify a hypothesis of close relationship.

Time banding

Vences et al. (2013) point out that taxa of the same rank do not necessarily represent comparable levels in a hierarchy and proposed some approaches towards reducing those disparities, in particular, the use of similar ranks at similar clade ages (Hennig 1966), a concept they refer to as time banding. They point out that time banding cannot be applied universally, but could be very useful within clades (e.g. families).

Accessory criteria

Entwisle & Weston (2005), Hopper (2009) and Vences et al. (2013) all took into account the disruption that name changes inflict on a broad community of users, even when well supported by evidence and leading to improved evolutionary interpretations. They advocate several similar accessory considerations that can help where criteria such as monophyly do not provide support for one or other of competing classifications. These include avoiding where possible classifications that change familiar names and circumscriptions of significant groups (e.g. Drosophila, Acacia) and seeking a consensus, or at least following a majority, of the taxonomic community.

Conclusion for this section

I support the following working criteria for genus recognition, in order of importance. In the remainder of this review, I apply these criteria to New Zealand's endemic genera.

1. Circumscriptions of named taxa can be objective if all are required to be monophyletic. Monophyletic taxa are objective, testable, accord with natural biological patterns and processes, facilitate the application of taxonomy in other pure and applied disciplines, and will, over time, lead to classifications that are stable. The requirement for monophyly involves ensuring not only that groups are monophyletic, but also that recognition of such groups does not render another group paraphyletic. In some cases, recognition of a distinctive clade in New Zealand (e.g. Colensoa, Embergeria, Haastia, Hebe, Kirkianella, Oreostylidium) has resulted in implied recognition of a grade in New Zealand or some other part of the world. Thus an international perspective is important.

2. Not all clades need to be named. Ideally, named clades should be recognisable based on morphological characters, and their monophyly should be strongly supported, preferably by independent lines of evidence. There may be circumstances in which named clades are not distinctive or distinctive clades remain unnamed.

3. Decisions about ranks for named taxa cannot be wholly objective, but the following principles can usefully be applied: (a) The phylogeny of the group should be represented in (and recoverable from) the classification where possible. (b) Clades recognised should be distinctive, but recognition of every distinctive clade is not necessary, especially if it mandates recognition of a non-distinctive sister clade or a distinctive grade. (c) Rank redundancy (named groups that have the same circumscription but different ranks) should be avoided where possible. (d) Existing and familiar names should be retained where possible consistent with the above guidelines.

Part 2. New Zealand endemic seed plant genera

The seed plant flora of New Zealand displays a high level of endemism at species rank. Godley (1975) published figures showing that 84.6% of the 1833 species recognised at that time were considered endemic. I recalculated endemism at species rank using the data in de Lange & Rolfe (2010): total endemism at species rank in New Zealand seed plants is 84.1% (Gymnosperms: 100%; Angiosperms 83.9%).

At family rank, only one endemic family, Ixerbaceae, has been recognised in recent years (Breitwieser et al. 2012), but the sister relationship of its sole genus Ixerba to New Caledonian Strasburgeria (Strasburgeriaceae; Sosa & Chase 2003; Matthews & Endress 2006; Carlquist 2007) and consequent placement in Strasburgeriaceae (Qiu et al. 2010) means that there are no endemic families currently recognised in New Zealand (de Lange & Rolfe 2010).

In contrast to high endemism at species rank, Millener (1960) and Godley (1975) noted that endemism at genus rank is low in New Zealand. Godley (1975) listed 38 endemic seed plant genera. A recent listing (Breitwieser et al. 2012) recognised 52 endemic Angiosperm genera, only 27 of which were also on Godley's list. An alternative (de Lange & Rolfe 2010) treated 58 genera as endemic. The changes since Godley (1975) reflect altered, and often less inclusive (Humphries & Linder 2009) circumscriptions of genera, especially in Orchidaceae (Clements & Jones 2002; Clements et al. 2002), and some altered knowledge of distribution ranges. Some genera, such as Iti and Heliohebe, have been proposed (Garnock-Jones & Johnson 1987; Garnock-Jones 1993b) and rejected (Heenan 2002; Garnock-Jones et al. 2007) in the period since Godley (1975).

List of endemic genera

The starting point for this review was a list of endemic seed plant genera drawn up from two recent checklists of the New Zealand Flora (de Lange & Rolfe 2010; Breitwieser et al. 2012) that also included information on endemism at genus rank. The two lists differ slightly, and genera considered endemic in one list but not the other were still included for consideration here. The combined list of potential endemic genera from these two sources (Table 1) comprises those genera currently recognised by New Zealand's largest plant taxonomy research community at Landcare Research (Breitwieser et al. 2012) for their herbarium and databases and also reflects the taxonomic opinions of New Zealand's field botanists and plant conservation community (de Lange & Rolfe 2010). A newer list (de Lange et al. 2013) does not address generic endemism explicitly, but its taxonomic deviations from de Lange & Rolfe (2010) have been addressed in Table 1 and elsewhere.

Table 1 New Zealand endemic seed-plant genera. The list combines recent lists of endemic genera (de Lange & Rolfe 2010 (*), Breitwieser et al. 2012 (†)). Names in bold are fully accepted as genera according to the criteria outlined above and are considered to have a distribution endemic to the political boundaries of New Zealand, although for some of these the notes suggest questions for future research. Underlined names are considered to be of doubtful taxonomic status (usually this means the monophyly of the genus or its close relatives is in question). All other names are rejected (either they have been shown not to be monophyletic, to render another genus paraphyletic, or they are distributed beyond political New Zealand).

Genus	Family	Monophyly	Monophyly of related groups	Clade diagnosability	Genus rank	Endemic to NZ (current circumscription)	Notes
Alepis*^†	Loranthaceae	+	?	+	?	+
Alseuosmia*^†	Alseuosmiaceae	+	+	+	+	+
Anemanthele*^†	Poaceae	+	Austrostipa possibly paraphyletic	+	?	+
Aporostylis*^†	Orchidaceae	+	+	+	+	+
Austroderia†	Poaceae	+	+	+	+	+
Clianthus*^†	Fabaceae	+	+	+	+	+
Colensoa*^†	Campanulaceae	+	Equivocal, appears to be nested within the Australian Lobelia sect. Holopogon	+	?	+
Dactylanthus*^†	Balanophoraceae	+	+	+	?	+	Combining it with Hachettia would remove rank redundancy
Damnamenia*^†	Asteraceae	+	+	+	?	+	Equivocal, could be combined with Pleurophyllum
Danhatchia*^†	Orchidaceae	+	+	+	+	Also in Australia
Dolichoglottis*^†	Asteraceae	+	+	+	+	+
Earina*	Orchidaceae	+	+	+	+	Also in South Pacific
Elingamita*^†	Primulaceae	+	+	+	+	+
Embergeria*^†	Asteraceae	+	= Sonchus	+	—	+
Entelea*^†	Malvaceae	+	+	+	+	+
Ewartiothamnus*	Asteraceae	+	Equivocal relationship to other NZ Gnaphalieae	+	?	+
Haastia*^†	Asteraceae	—	= Brachyglottis	+	—	+
Halocarpus*^†	Podocarpaceae	+	+	+	+	+
Hectorella*^†	Montiaceae	+	+	+	?	+	Combining it with Lyallia would remove rank redundancy
Heliohebe*	Plantaginaceae	+	= Veronica	+	—	+
Hoheria*^†	Malvaceae	+	+	+	+	+
Ichthyostomum*	Orchidaceae	+		?	—	+	= Bulbophyllum
Ixerba*^†	Strasburgeriaceae	+	+	+	+	+	Combining it with Strasburgeria would remove rank redundancy at genus rank but introduce it at family rank.
Kirkianella*^†	Asteraceae	+	= Sonchus	+	—	+
Knightia*^†	Proteaceae	+	+	+	+	+
Leonohebe*	Plantaginaceae	+	= Veronica	+	—	+
Leucogenes*^†	Asteraceae	? (morphological but not molecular support)	?	+	?	+	Equivocal, part of Raoulia complex
Lignocarpa*^†	Apiaceae	+	Equivocal, part of Anisotome complex.	+	?	+
Lophomyrtus*^†	Myrtaceae	?	+	+	?	+
Manoao*^†	Podocarpaceae	+	+	+	?	+	Equivocal; I prefer to incude it in Lagarostrobos
Mida*^†	Nanodeaceae	+	+	+	?	+	Combining it with Nanodea would remove rank redundancy at genus rank but introduce it at family rank
Molloybas*^†	Orchidaceae	+	+	+	?	+	Equivocal, = Corybas?
Montigena*^†	Fabaceae	+	= Swainsona	+	—	+
Myosotidium*^†	Boraginaceae	+	current circumscription of Omphalodes paraphyletic	+	?	+
Neomyrtus*^†	Myrtaceae	+	Equivocal, = Lophomyrtus?	+	?	+
Notothlaspi*^†	Brassicaceae	+	+	+	+	+
Oreostylidium*^†	Stylidiaceae	+	= Stylidium	+	—	+
Pachystegia*^†	Asteraceae	+	Equivocal, relationship to Shawia clade (Olearia p.p.) needs clarification	+	?	+
Peraxilla*^†	Loranthaceae	?	Equivocal, = Alepis?	+	?	+
Plagianthus*^†	Malvaceae	+	+	+	+	+
Pleurophyllum*	Asteraceae	+	Current circumscription of Olearia is non-monophyletic	+	?	Also on Macquarie Island
Pseudowintera*^†	Winteraceae	+	+	+	+	+
Pyrrhanthera*^†	Poaceae	+	= Rytidosperma	+	—	+	de Lange (2011a) rescinded acceptance of Pyrrhanthera
Rachelia*^†	Asteraceae	+	?	+	?	+	Equivocal, part of Raoulia complex.
Raoulia*^†	Asteraceae	? No molecular support, morphology equivocal	Monophyly of related genera uncertain	+	?	+
Rhabdothamnus*^†	Gesneriaceae	+	+	+	+	+
Scandia*^†	Apiaceae	+	Equivocal, part of Anisotome complex	+	?	+
Simplicia*^†	Poaceae	+	+	+	+	+
Singularybas*^†	Orchidaceae	+	+	+	?	+	Equivocal, = Corybas?
Stenostachys*^†	Poaceae	+	+	+	+	+
Stilbocarpa*	Apiaceae	+	Related Schizeilema and Huanaca non-monophyletic	+	?	Also on Macquarie Island
Teucridium*^†	Lamiaceae	+	+	+	+	+	Monophyly of Teucrium needs to be established
Toronia*^†	Proteaceae	+	Monophyly of Persoonia not established	+	?	+
Traversia*^†	Asteraceae	+	= Brachyglottis?	+	—	+
Trilepidea*^†	Loranthaceae	+	Monophyly of Peraxilla and Alepis needs examination	+	?	+	Relationship to Alepis needs reconsideration.
Tupeia*^†	Loranthaceae	+	+	+	+	+
Waireia*^†	Orchidaceae	+	+	+	+	+
Winika*^†	Orchidaceae	+	= Dendrobium?	+	?	+
Zotovia*^†	Poaceae	+	Ehrharta non-monophyletic	+	—	+

Nomenclature

The assessment of the status of New Zealand's endemic genera presented below results in some clear recommendations about taxonomy. I present a new combination in only one such case, Sonchus novae-zelandiae. For others (e.g. Rytidosperma exiguum, Sonchus grandifolius, Stylidium subulatum) binomials are already available. In other endemic genera, taxonomic status is still equivocal and awaits the conclusion of current research projects (e.g. in the Raoulia complex, I. Breitwieser pers. comm.). For a number of genera (e.g. Teucridium, Toronia, Trilepidea), there is so far insufficient research that bears on their monophyly or that of their close relatives, and so although I recommend provisional taxonomic acceptance, I indicate in the discussion that more research is needed.

Gymnosperms

Podocarpaceae

Halocarpus Quinn (3 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

The partition of traditional large genera of Podocarpaceae, Dacrydium and Podocarpus (de Laubenfels 1969; Quinn 1982), seems justified by the monophyly of morphologically divergent subclades such as Halocarpus (Wagstaff 2004) and by the great age of these groupings. Biffin et al. (2011) used matK, trnL–F and internal transcribed spacer (ITS)2 sequence data to show the three species of Halocarpus form a well-supported clade that renders no other segregate genus paraphyletic. Burleigh et al. (2012) presented a supertree, derived from previous mitochondrial, chloroplast and nuclear DNA studies, which supports monophyly of Halocarpus (Fig. 2). Morphology also strongly supports the relationship of the three New Zealand species assigned to Halocarpus. Divergence of Halocarpus was dated at (112–) 87 (–61) million years old by Biffin et al. (2011), indicating a likely Gondwanan vicariance origin. I accept Halocarpus as a genus endemic to New Zealand.

Figure 2 Relationships within some Podocarpaceae, abridged from a maximum likelihood supertree constructed by Burleigh et al. (2012), showing sister relationship of Manoao and Lagarostrobos, Halocarpus as a distinct and monophyletic lineage, and also the paraphyly of Podocarpaceae if Phyllocladus and Prumnopitys are excluded. Estimated ages of branching events also shown.

Manoao Molloy (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Manoao is a monospecific genus segregated by Molloy (1995) from Lagarostrobos, consequently also monospecific. Molloy (1995) documented morphological differences between M. colensoi and L. franklinii of Tasmania, and argued that their separation at generic rank is justified by the extent of their differences. A sister relationship between Manoao and Lagarostrobos has been demonstrated with 18S rDNA (Kelch 2002; moderate support), matK, trnL–F and ITS2 (Biffin et al. 2011; strong support), rbcL (Knopf et al. 2012; strong support), and in a supertree presented by Burleigh et al. (2012; Fig. 2), but using the NEEDLY intron 2, Manoao was weakly supported as sister to Halocarpus (Knopf et al. 2012). Thus, although most evidence supports the monophyly of Lagarostrobos (both s.l. and s.s.) and Manoao, at least with respect to extant species, there is conflict between two competing criteria for deciding rank. On the one hand, recognising two monospecific genera that are sisters makes the genus rank redundant and the classification is thus nowhere informative about the sister relationship between them (Kelch 2002). Kelch (2002) concluded ‘A conservative approach (i.e. one avoiding unnecessary monotypes) favours the retention of L. colensoi in Lagarostrobos, pending further evidence of relationships within the group’. On the other hand, segregation of Manoao does appear to be widely accepted in New Zealand (de Lange & Rolfe 2010; Dawson & Lucas 2011; Breitwieser et al. 2012; but see Miller 2006) and the inferred late Cretaceous age of the separation of L. franklinii and M. colensoi is similar to that of other generic separations in the family (Burleigh et al. 2012; but see Biffin et al. 2011 who dated that split as Paleogene). Christenhusz et al. (2011) accepted Manoao without discussion and Biffin et al. (2011) concluded that the morphological differences between Manoao and Lagarostrobos justify their continued recognition. Jordan et al. (2011) identified Oligocene/Miocene pollen from New Zealand as Lagarostrobos and noted that Manoao is not known in the fossil record. This suggests the possibility that Lagarostrobos would be paraphyletic if extinct species were included. In my opinion, classifying L. colensoi in the same genus as its closest living relative is preferable to a classification that reflects its phenetic distance from L. franklinii.

Angiosperms

Alseuosmiaceae

Alseuosmia A.Cunn (5 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

The monophyly of Alseuosmia has not been questioned. The genus is strongly supported as sister to Wittsteinia and more distantly related to Crispiloba and Platyspermation (Tank & Donoghue 2010 and references therein), based on extensive sampling of the chloroplast genome and nuclear ribosomal DNA. More extensive taxon sampling is needed to confirm monophyly, but provisionally I accept Alseuosmia as a genus endemic to New Zealand.

Apiaceae

Lignocarpa J.W.Dawson (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Scandia J.W.Dawson (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Lignocarpa and Scandia were segregated from Anisotome (Dawson 1967), initially as Gingidium p.p. (Dawson 1961). Lignocarpa (two species) is a morphologically divergent clade of scree-adapted plants. Scandia (two species) is a clade of softly woody semi-climbing plants. Together with non-endemic Aciphylla, Anisotome and Gingidia, they make up the Aciphylleae (Downie et al. 2010). This complex is a clade of morphologically highly divergent plants adapted to differing environments, a classic adaptive radiation in New Zealand on a similar scale to those of Brachyglottis, Celmisia and Veronica.

Mitchell et al. (1998), Radford et al. (2001) and Spalik et al. (2010) found congruent results for this group using ITS sequence data (Fig. 3A). Gingidia decipiens and G. flabellata are nested within Anisotome, whereas the remaining species of Gingidia form a clade with Scandia and Lignocarpa that is nested within Aciphylla. Webb & Druce (1984) described an intergeneric hybrid, Aciphylla squarrosa × Gingidia montana, which is supporting evidence of a close relationship. Furthermore the Aciphylla–Gingidia–Lignocarpa–Scandia clade is nested within paraphyletic Anisotome when large numbers of Aciphylla and Anisotome are included (Radford et al. 2001).

Figure 3 Relationships in Apiaceae. A, Supertree of Apioideae constructed by inspection (after ITS trees of Mitchell et al. 1998; Radford et al. 2001; Spalik et al. 2010; these three studies were congruent except that Mitchell et al. attached Aciphylla dieffenbachii in a polytomy at the basal node). B, Apioideae (after Nicolas & Plunkett 2009—cpDNA: trnD–trnT, rpl16; BS, PP values > 50% shown; +, BS between 80 and 94%; *, BS > 95%). C, Relationships of Stilbocarpa shown as a summarised topology recovered with maximum likelihood and Bayesian analyses of cpDNA (rpl16, trnD–trnT, PP not shown, but < 1.0; *, branches with low support) by Nicolas & Plunkett 2012.

These analyses suggest the recognition of at best two genera, Anisotome (including G. flabellata and G. decipiens) and Aciphylla (including the remaining species of Gingidia, Anisotome procumbens, Lignocarpa and Scandia). Given the morphological distinctiveness of Aciphylla, Scandia and Lignocarpa, this is likely to be contentious, yet classifying them together draws attention to this spectacular adaptive radiation and puts evolutionary studies, such as the detailed descriptions and evolutionary pathways of sexual systems provided by Webb (1979), in a firm phylogenetic framework. Before making the nomenclatural changes, wider sampling of species (especially in Anisotome) and genes is recommended. Caution is also indicated by the study of Nicolas & Plunkett (2009). They used chloroplast DNA only and did not find the same topology as Radford et al. (2001; Fig. 3B); here a clade of Scandia, Lignocarpa and Gingidia (one species each) was well supported as a sister to a clade that combined six species of Aciphylla and Anisotome.

Stilbocarpa (Hook.f.) Decne. & Planch. (3 spp.; de Lange & Rolfe 2010)

De Lange & Rolfe included Stilbocarpa as endemic because they included Macquarie Island in the New Zealand botanical region (see also Allan 1961). That is not the convention followed here and by Breitwieser et al. (2012), for which endemism to the political region is the criterion.

Nicolas & Plunkett (2009, 2012) showed that Stilbocarpa is nested within an otherwise South American clade containing the genus Huanaca and one species of Schizeilema (Fig. 3C). Rearrangements of generic boundaries are needed, so that Schizeilema and Huanaca can have monophyletic circumscriptions; if treated as a single genus, the oldest genus name for the Huanaca/Stilbocarpa clade appears to be Huanaca.

Asteraceae

Asteraceae includes several large independent adaptive radiations (treated separately below) where rapid and relatively recent adaptation to different habitats has driven morphological divergence, often extreme, within clades. This in turn has encouraged recognition of distinctive small clades as genera separate from the distinctive large grades they are nested within.

1. Celmisia–Olearia complex (Tribe Astereae)

This generic complex comprises a clade that includes species currently referred to Celmisia, Damnamenia, Olearia p.p., Pachystegia and Pleurophyllum, and an additional clade that includes Australian species that are more distantly related (Olearia A + B of Cross et al. 2002, which includes the type of the genus). Three of these genera—Damnamenia, Pachystegia and Pleurophyllum—have been listed as endemic in recent accounts.

Damnamenia Given. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012), Pachystegia Cheeseman (3 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Pleurophyllum Hook.f. (3 spp.; de Lange & Rolfe 2010)

Taxonomic resolution of the Celmisia–Olearia complex into monophyletic groups involves broad considerations of Olearia, Celmisia and many smaller genera throughout their ranges. It will likely require further collaborative study that includes investigators and samples from Australia, New Zealand and South America.

Recent molecular phylogenetic research (Cross et al. 2002; Wagstaff et al. 2011; Fig. 4) provides evidence that Celmisia, Pleurophyllum, Damnamenia and Pachystegia are each monophyletic as currently circumscribed. However, Olearia is not monophyletic; Cross et al. (2002) include species of Olearia in five major clades (they identified several smaller clades within these most inclusive ones). In that study, which analysed ITS sequences, an Australian clade (which includes the type species of Olearia as well as samples of Oritrophium and Felicia) is sister to a clade of Olearia from Australia and New Zealand (including the type species of Shawia), Celmisia, Damnamenia, Pleurophyllum and Pachystegia. Wagstaff et al. (2011) used ITS, external transcribed spacer (ETS) and two chloroplast sequences to further analyse this mostly New Zealand clade. In their combined analysis, Celmisia was monophyletic and sister to a clade comprising Olearia p.p. (Shawia clade) and its sister group Pachystegia. Both studies recovered a clade comprising Damnamenia, Pleurophyllum and the six species of macrocephalous Olearia as sister to a clade comprising Olearia p.p. (Shawia), Pachystegia and Celmisia (Fig. 4).

Figure 4 Relationships in Astereae shown as a simplified tree of the largely New Zealand Olearia–Celmisia complex (ITS + ETS + trnK + trnL; PP = 1 unless shown; Wagstaff et al. 2011) in the wider context of the ITS tree of Cross et al. (2002).

Damnamenia. Given (1973) used largely phenetic criteria to raise Celmisia sect. Antarcticae (Allan 1961) to genus rank as Damnamenia, arguing its one species lacked character states characteristic of mainland Celmisia, yet shared a number with Pleurophyllum and macrocephalous Olearia. The phenetic similarity of Damnamenia with Pleurophyllum and macrocephalous Olearia (Drury 1968; Given 1973) suggested relationships that have since been supported by explicitly phylogenetic methods and molecular data (Wagstaff et al. 2011). The three species of Pleurophyllum are sister to a clade of six species of macrocephalous Olearia, with the monospecific Damnamenia as sister to these nine species (Cross et al. 2002; Wagstaff et al. 2011). The distinctiveness of these mostly southern species of macrocephalous Olearia has long been recognised (Allan 1961) and Drury (1968) was first to ally them with Pleurophyllum.

There are several possible taxonomic solutions for this grouping following the criteria recommended here, two of which are likely outcomes. First, describing a new genus for macrocephalous Olearia would have the advantage of leaving the nomenclature of Pleurophyllum and Damnamenia unchanged. Second, Wagstaff et al. (2011) argued that Damnamenia could be combined with Pleurophyllum and macrocephalous Olearia as a single near-endemic genus, for which the correct name would be Pleurophyllum. Thus Pleurophyllum would be enlarged to include Damnamenia, Olearia angustifolia, O. oporina, O. chathamica, O. semidentata, O. lyallii and O. colensoi. This wider concept of Pleurophyllum is supported by a hybrid between Damnamenia vernicosa and Pleurophyllum speciosum (as Celmisia ×campbellensis F.R.Chapman = C. ×chapmanii T.Kirk), species separated at the most basal node in the clade. Allan (1961) and Given (1973) referred to a hybrid plant found by W.B. Brockie on Mt Lyall, additional to Chapman's collection from about 12 plants seen at Venus Bay. On current evidence, I prefer the latter solution.

Pleurophyllum

Whatever its eventual circumscription, Pleurophyllum is not endemic to New Zealand because P. hookeri extends to Macquarie Island (Breitwieser et al. 2012), although Allan (1961) and de Lange & Rolfe (2010) included Macquarie Island in the New Zealand botanical region.

Pachystegia

This genus is weakly supported either as sister to a clade of species from Australia and New Zealand (Olearia clades E + H of Cross et al. 2002) or sister to a larger clade of Olearia (equivalent to clades E, F, G and H of Cross et al. 2002) (e.g. Wagstaff et al. 2011). Current evidence is thus equivocal, and future acceptance of Pachystegia will depend on the outcome of more detailed study of Olearia and resolution of the conflicts between the phylogenetic signals from ITS and ETS (Wagstaff et al. 2011).

Based on current limited sampling, Celmisia is monophyletic and either sister to a clade that includes O. paniculata, type species of Shawia Forst.f. (Wagstaff et al. 2011), or nested within a larger grade of Australian and New Zealand Olearia (Cross et al. 2002). The topology needs to be resolved if a monophyletic Shawia is to be circumscribed and wider sampling of Celmisia will be needed before its monophyly can be confirmed. Note also an additional intergeneric hybrid, between Olearia arborescens (Shawia clade) and Celmisia gracilenta (Clarkson 1988).

2. Brachyglottis complex (Tribe Senecioneae)

Dolichoglottis B.Nord. (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012), Haastia Hook.f. (3 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Traversia Hook.f. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Nordenstam (1978) extended the circumscription of Brachyglottis from a single species (Allan 1961) to include 30 species treated under Senecio by Allan (1961), at the same time segregating Dolichoglottis (two species) and Urostemon (one species, transferred to Brachyglottis by Webb in Connor & Edgar 1987). For this grouping of Senecioid genera, Pelser et al. (2007) using ITS and chloroplast DNA (cpDNA), recovered a well-supported Caputia–Brachyglottis clade (as Senecio medley-woodii–Brachyglottis clade: see Nordenstam & Pelser 2012), in which the current circumscription of Brachyglottis is paraphyletic (Fig. 5). The segregate genera Acrisione, Urostemon, Papuacalia, Traversia, Centropappus, Bedfordia and Haastia are all nested within Brachyglottis. Additionally, Dolichoglottis is placed either as sister to Brachyglottis (Pelser et al. 2007) or nested within it (Wagstaff & Breitwieser 2004). Of these, Dolichoglottis, Haastia and Traversia are considered endemic to New Zealand.

Figure 5 Relationships in the Brachyglottis complex (ITS tree, Bayesian consensus probability = 100 unless shown; Pelser et al. 2007).

Haastia was traditionally classified in either Tribe Gnaphalieae or Tribe Astereae until Breitwieser & Ward (2005) transferred it to Senecioneae based on morphology (Breitwieser & Ward 2005), anatomy and flavonoid chemistry (Breitwieser & Ward 1993; Breitwieser 2008), palynology (Breitwieser & Sampson 1997a,b) and molecular systematics (Wagstaff & Breitwieser 2002, 2004). Its phylogenetic position, nested within Brachyglottis, and the lack of evidence for its monophyly have both been supported in more recent studies (e.g. Pelser et al. 2007; Fig. 5).

Making generic circumscriptions monophyletic in this complex will have interesting nomenclatural consequences. On current evidence, if Haastia were to be retained with the current species included, it would need to be enlarged to also include several leafy shrubs such as Brachyglottis perdicioides and B. monroi. In that scenario, Brachyglottis would revert to its earlier, narrow circumscription (Allan 1961). Otherwise, evidence supports the inclusion of Haastia within Brachyglottis. This solution would draw attention to the close relationship of these alpine cushion plants to shrubs from lower altitude, an adaptive radiation that parallels the snow hebes within Veronica, the vegetable sheep within the Raoulia complex and pulvinate Myosotis.

At the basal nodes of the Brachyglottis clade, support values for many branches are moderate at best (Pelser et al. 2007), indicating that a broad circumscription of Brachyglottis is sensible. This would enlarge the genus to include not only Haastia, but also Acrisione, Urostemon, Papuacalia, Traversia, Centropappus and Bedfordia. In that circumscription, Dolichoglottis could be accepted as a genus endemic to New Zealand. A second option would be to enlarge the circumscription of Brachyglottis a little further to include Dolichoglottis. A third option would divide the complex narrowly to contain five New Zealand endemic genera, but there is currently only weak support for some of these clades (Pelser et al. 2007). I accept Dolichoglottis, but not Haastia and Traversia, as genera endemic to New Zealand

3. Sonchus complex (Tribe Lactuceae)

Embergeria Boulos (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Kirkianella Allan (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Kim et al. (2007) used ITS and cpDNA data to show the Sonchus alliance includes many small, often island, genera (including Embergeria and Kirkianella) that are nested within a paraphyletic or polyphyletic circumscription of Sonchus (Fig. 6). They recommended that the circumscription of Sonchus be enlarged to include these segregates and avoid breaking up that largely uniform genus. Mejías & Kim (2012) transferred Dendroseris and Thamnoseris to Sonchus.

Figure 6 Relationships in the Sonchus complex (consensus tree of combined ITS and matK + flanking sequences; bootstrap support shown; Kim et al. 2007).

Embergeria is clearly nested within Sonchus, based on evidence from molecular data (Kim et al. 2007 and references therein). Apart from its larger stature and large fruits, the absence of stiff clavate barbellate bristles among the shorter slender soft pappus hairs appears to be an apomorphy perhaps associated with loss of dispersal in its island habitat (Garnock-Jones in Webb et al. 1988). The correct name in Sonchus for its single species is S. grandifolius Kirk. Within Sonchus it appears most closely related to S. novae-zelandiae (Kirkianella novae-zelandiae, see below); both have only soft fine pappus hairs.

Inclusion of Kirkianella within Sonchus is supported based on evidence from molecular data (Kim et al. 2007 and references therein) and morphology (Bentham & Hooker 1873). Like S. grandifolius, it appears to lack stiff barbellate pappus bristles, which supports the molecular finding (Kim et al. 2007) of a sister relationship between them. Allan (1961) segregated the genus from its previous placement in Crepis based on its solitary capitula. Although the International Plant Names Index (www.ipni.org) cites Bentham & Hooker (1873) as making the combination in Sonchus, they did not definitely associate the epithet with the genus name (ICN, Art. 35.2), but instead only gave their opinion of where the species should be classified.

Sonchus novae-zelandiae (Hook.f.) Garn.-Jones, comb. nov. ≡ Crepis novae-zelandiae Hook.f., Handbook of the New Zealand Flora: 164 (1864). ≡ Kirkianella novae-zelandiae (Hook.f.) Allan, Flora of New Zealand 1: 762 (1961). Lectotype: S. novae-zelandiae has not previously been formally lectotypified, although Allan (1961, p. 762) recommended a Haast specimen at K should be selected. However the Haast specimens in Hooker's herbarium (at K) lack fruits, which were clearly described in the protologue, and are thus not as representative as others. Four collections were listed in the protologue, collected by Banks & Solander, Sinclair & Haast, Lindsay, and Hector & Buchanan. The last three are mounted together on one sheet in Hooker's herbarium at Kew. The Banks and Solander material is at BM, WELT, and AK. I designate the Banks & Solander material at BM as lectotype because it includes all the life history stages described in the protologue; additionally it is a high quality collection and duplicates are available in New Zealand (WELT, AK).

The S. novae-zelandiae complex is morphologically diverse and cytologically variable and revision at species rank is recommended (Allan 1961; Beuzenberg & Hair 1984; Dawson 2000; de Lange & Rolfe 2010; de Lange et al. 2013).

4. Raoulia alliance (Tribe Gnaphalieae)

Ewartiothamnus Anderb. (1 sp.; de Lange & Rolfe 2010), Leucogenes Beauverd (4 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012), Rachelia J.M.Ward & Breitw. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Raoulia Hook.f. (23 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

The Raoulia alliance comprises a distinctive grouping of plants that includes a range of distinct morphological (e.g. cushion plants, mat plants, leafy and whipcord shrubs) and anatomical types (Breitwieser & Sampson 1997a, b; Breitwieser & Ward 1998). Based on current circumscriptions, several hybrids are treated as intergeneric within the Raoulia alliance (Ward 1997; McKenzie et al. 2004; Smissen et al. 2007), indicating close genetic compatibility in spite of the divergent morphologies (Morgan-Richards et al. 2009). Monophyly of the current generic circumscriptions has not been convincingly demonstrated with molecular evidence (Fig. 7), probably because of this complex's history of adaptive radiation, hybridism and reticulate evolution, and divergent and possibly convergent morphological evolution (Smissen et al. 2011). Including all groupings (currently genera, most endemic) within one morphologically variable but genetically closely related and probably endemic genus is just one of several possible solutions, but one that would recognise the close relationship among these distinctive New Zealand plants within the wider classification of the tribe. However, that step would be premature while current research is investigating the complex evolution of the group (Breitwieser et al. 1999; Smissen et al. 2007, 2011). I have reflected this uncertainty here by accepting Raoulia with an uncertain circumscription, while labelling the other genera in the Raoulia complex as taxonomically equivocal (Table 1).

Figure 7 Relationships in the Raoulia alliance. A, Nuclear DNA tree based on ITS and ETS; PP shown if > 0.95. B, Chloroplast DNA tree based on trnL intron, trnL–F, ndhF, trnK intron [incl. matK]; PP shown if > 0.95 (simplified from Smissen et al. 2011).

Balanophoraceae

Dactylanthus Hook.f. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Nickrent (http://www.parasiticplants.siu.edu/ListParasites.html, accessed 27 September 2013) accepted Dactylanthus as a monospecific genus endemic to New Zealand. A sister relationship between Dactylanthus and Hachettia of New Caledonia, also monospecific, has been suggested (see Fay et al. 2010). I note that very small genera seem to be a feature of the classifications of parasitic plants throughout the world. I accept Dactylanthus as a genus endemic to New Zealand.

Boraginaceae

Myosotidium Hook. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Heenan et al. (2010) obtained ambiguous indications of the relationships of Myosotidium hortensia¹ in a NeighborNet comparison of DNA sequence similarities. On the one hand, analysis of ITS sequences indicated the genus has similar sequences to Omphalodes, a possible relationship previously alluded to by Hooker (1859) on morphological grounds. On the other hand, its chloroplast atpB sequences were not closely similar to those of other genera, although Heenan et al. (2010) discussed sequence similarities with Lappula and Trichodesma.

Nazaire & Hufford (2012; Fig. 8A) used data from the ITS region of nuclear ribosomal DNA and ndhF, matK, rbcL and trnL–F from the chloroplast genome in a phylogenetic analysis of Boraginaceae. Khoshsokhan et al. (2013) included M. hortensia in a study of Boraginaceae trib. Eritrichieae using nuclear ITS data. Cohen (2013; Fig. 8B) used ITS and three chloroplast sequences (matK, ndhF and trnL–trnF). Their results are broadly similar to those of Heenan et al. (2010) and most analyses suggest the current circumscription of Omphalodes is paraphyletic unless Myosotidium is included in its circumscription. Cohen (2013) also included an analysis and discussion of morphological characters and showed that Myosotidium shares three synapomorphies with Omphalodes: secondary leaf venation (a labile character state in the family), cordate leaves and nutlets with marginal glochids or wings. Cohen recommended wider taxon sampling within Omphalodes to resolve the taxonomy within this clade. These four largely congruent results strongly indicate that even if Myosotidium is retained at genus rank it is very likely to have a wider and no longer endemic circumscription.

Figure 8 Relationships of Myosotidium (Boraginaceae). A, Maximum likelihood analysis of combined data from the ITS region and four chloroplast sequences (matK, ndhF, rbcL and trnL–F). ML bootstrap/BI posterior probabilities shown. From Nazaire & Hufford (2012; Fig. 4). B, A strict consensus maximum parsimony tree of combined molecular (ITS, matK, ndhF, trnL–trnF) and morphological data sets, jackknife values shown (Cohen 2013).

Brassicaceae

Notothlaspi Hook.f. (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Notothlaspi is a distinctive clade, for which close relatives have yet to be established. Some of its distinctive features, such as the thick grey–green leaves and disproportionately long radicle (Hooker 1864; Garnock-Jones 1991a) are likely to be adaptations to its scree and stonefield habitats, and might obscure its relationships to other genera. However, it is apparent that its nearest relatives are not in New Zealand (Mitchell & Heenan 2000). Recent work (Warwick et al. 2010; Al-Shehbaz 2012) placed it in its own tribe, Notothlaspideae, closest among the genera sampled to the South African Heliophila, Trib. Heliophileae (Fig. 9). I accept Notothlaspi as a genus endemic to New Zealand.

Figure 9 Relationships of Notothlaspi (Brassicaceae) and its closest relatives among genera sampled so far (from Warwick et al. 2010, strict consensus ITS; parsimony bootstrap values shown).

Campanulaceae

Colensoa Hook.f. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

The genus includes only one species, Colensoa physaloides, treated as Pratia physaloides by Allan (1961). Botanists, including in New Zealand, generally are adopting a wider circumscription of Lobelia, to include Pratia, Isotoma and Hypsela (Antonelli 2008; Heenan et al. 2008; Knox et al. 2008; Lammers 2011). However, while the transfers of Pratia, Isotoma and Hypsela to Lobelia have been generally accepted, many botanists (e.g. de Lange & Rolfe 2010; Breitwieser et al. 2012; E.B. Knox, pers. comm.) continue to treat Colensoa as a separate genus. There is good evidence (Antonelli 2008) that Colensoa physaloides (as Lobelia physaloides) is not closely related to the other New Zealand species of Lobelia, including those formerly classified with it under Pratia. Nevertheless, Colensoa is nested within the current circumscription of Lobelia s.l., which thus appears to be paraphyletic. Overseas, the genus is very variable in habit, flower size and shape, and fruit structure. Many characters are homoplasious, and additionally, L. physaloides has significant autapomorphies (Lammers 2011).

Lammers (2011; following Antonelli 2008) provided a taxonomic treatment, in which L. physaloides was placed on its own in Lobelia sect. Colensoa, related to a large section of species (sect. Delostemon) that is found in Africa, Asia and South America and perhaps also to a smaller section (sect. Holopogon) that is endemic to Australia. Thus, if Lobelia is not to be expanded to have an identical circumscription to the subfamily Lobelioideae, L. physaloides belongs among a group of species with a good claim to generic rank (Clade C1 of Antonelli 2008 or some part of it, Fig. 10). Since that study, E.B. Knox (pers. comm.) has increased taxon sampling in the group by including species of the Australian clade Lobelia sect. Holopogon, within which he found L. physaloides is nested.

Figure 10 Relationships in Campanulaceae subfam. Lobelioideae after Antonelli (2008: trnL–F region; rbcL, ndhF; PP shown where < 1.0, not shown for C1 clade). C1 clade modified following EB Knox (pers. comm.) by addition of L. sect. Holopogonae.

If Colensoa is to be retained as a genus, its circumscription should be reconsidered. To maintain monophyletic groups, related species could be treated as congeneric with Colensoa and the C1 clade could be treated at generic rank or further subdivided. However, only by very fine division of the C1 clade could Colensoa remain as a New Zealand endemic. Expanding the circumscription of Lobelia to have the same circumscription as subfam. Lobelioideae would be a less informative alternative.

Clianthus Lindl. (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Montigena Heenan. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Both Clianthus and Montigena are related to Carmichaelia. Clianthus is sister clade to Carmichaelia (Heenan 1998a; Wagstaff et al. 1999; Fig. 11), distinguished from it by its leafiness, large red flowers and large, many-seeded follicular fruits. Both genera appear to have apomorphic characters and there are substantial morphological differences between them. Maintaining Clianthus separate from Carmichaelia seems acceptable, and taxonomic recognition of their sister relationship at a higher rank than genus may be justified.

Figure 11 Relationships of Carmichaelia, Clianthus, Montigena and Swainsona (Fabaceae); majority rule ITS tree of Wagstaff et al. (1999; support values not shown).

Heenan (1998b) proposed the new genus Montigena to reflect significant morphological and anatomical differences from Swainsona, yet the strong support from molecular data for its inclusion in Swainsona suggests such features are autapomorphies and adaptations to its New Zealand alpine habitat. Thus, recognition of Montigena is not recommended.

Recognition of Clianthus, Carmichaelia and Montigena renders the Australian genus Swainsona paraphyletic. One clade within Swainsona (including S. formosa) is sister to a clade comprising Carmichaelia, Clianthus and the rest of Swainsona. Resolution of that situation is perhaps more likely to lead to subdivision of Swainsona than to synonymy of Carmichaelia.

Gesneriaceae

Rhabdothamnus A.Cunn. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

According to a phylogeny based on ITS, G-CYCLOIDEA and four cpDNA sequences (Woo et al. 2011), this monotypic endemic genus is sister to Coronanthera, which has up to 20 species in New Caledonia and one in Papua New Guinea and the Solomon Islands. Rhabdothamnus, a twiggy shrub with solitary flowers and showy orange tubular corollas with spreading lobes, is also morphologically distinct from Coronanthera, trees with cymose inflorescences and small white, green, yellow, or brownish campanulate small-lobed corollas (Woo et al. 2011). I accept Rhabdothamnus as a genus endemic to New Zealand.

Lamiaceae

Teucridium Hook.f. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

A molecular study based on rbcL and ndhF sequence data (Wagstaff et al. 1998) placed Teucridium as sister to the monospecific genus Oncinocalyx of southeast Australia (Fig. 12). These two small genera were shown to be sister to Teucrium fruticans, the sole representative in that study of a large and widespread genus, a relationship supported also by pollen (Abu-Asab & Cantino 1993) and chemical (Grayer et al. 2002) data. Although there is no indication that Teucridium is nested within Teucrium, that possibility has not been refuted by including a larger sample of species. Monophyly of Teucrium needs to be established before Teucridium can be unequivocally accepted.

Figure 12 Relationships of Teucridium (Lamiaceae; Wagstaff et al. 1998, rbcL and ndhF; bootstrap values shown).

Loranthaceae

Alepis Tiegh. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012), Peraxilla Tiegh. (2 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Trilepidea Tiegh. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Allan (1961) included Alepis, Peraxilla and Trilepidea in a broad circumscription of Elytranthe, following Engler (1897). A sister relationship between Alepis flavida and Peraxilla tetrapetala is well-supported and these are sister to Elytranthinae (Vidal-Russell & Nickrent 2008; Fig. 13) based on a phylogenetic analysis of chloroplast and nuclear DNA sequence data. However, because only one of the two species of Peraxilla was sampled, its monophyly has yet to be demonstrated.

Figure 13 A summary of relationships within Loranthaceae (modified from Vidal-Russel & Nickrent 2008) based on data from matK, trnL–F, rbcL and SSU rDNA; PP shown for clades.

The chief qualitative characters that separate Alepis and Peraxilla (Alepis: corolla gamopetalous; anthers dorsifixed near the base; Peraxilla: corolla choripetalous; anthers basifixed) are not necessarily substantial and might relate to changes in pollinators (Vidal-Russell & Nickrent 2008). Furthermore, even if Peraxilla is shown to be monophyletic, its recognition as a genus distinct from Alepis obscures their close relationship; combining them would probably make a more informative classification and nomenclature.

Trilepidea is a monospecific genus and presumed extinct (de Lange et al. 2010); thus it has not been included in molecular phylogenetic studies, but was included in Tribe Elytrantheae by Vidal-Russell & Nickrent (2008). It appears similar to Alepis and Peraxilla, differing chiefly in its 3 persistent floral bracts, 5–6-merous corolla and apiculate anthers. Of these, I suspect the first and last are likely to be autapomorphies, whereas the 5–6-merous corolla is likely to be plesiomorphic. The additional floral bracts might be related to a reduction in inflorescence flower number in Trilepidea. Thus, although different, there may be little evidence to support its generic distinction from Alepis and Peraxilla. Currently, there is not evidence that falsifies the hypothesis that Alepis, Peraxilla and Trilepidea are a single adaptive radiation in New Zealand and more research is needed before three endemic genera can be accepted with confidence.

Tupeia Cham. & Schltdl. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Tupeia and Ileostylus are both monospecific and are separate lineages from each other and from Alepis, Peraxilla and Trilepidea (Vidal-Russell & Nickrent 2008; Fig. 13). Their relationships appear to be with South American and Australian plants (Desmaria and Muellerina) respectively (Vidal-Russell & Nickrent 2008), and their continued generic status depends on comparisons beyond New Zealand. I accept Tupeia as a genus endemic to New Zealand, whereas Ileostylus lost its endemic status when it apparently self-introduced to Norfolk Island in the 1930s (Mills 2010).

Malvaceae

Entelea R.Br. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Phylogenetic analysis of ndhF sequence data (cpDNA) strongly supported a sister relationship of Entelea arborescens with Sparrmannia africana and a more distant relationship between those two and Apeiba tibourbou (Brunken & Muellner 2012). In a parsimony analysis of morphological data, which included additional genera, Entelea and Clappertonia were weakly supported as a clade sister to Sparrmannia (Brunken & Muellner 2012). Currently, Sparrmannia is a genus of seven species, characterised by pendent flowers with a recurved corolla and irritable stamens (Mabberley 2008). These results are consistent with less conclusive earlier findings based on morphological (Judd & Manchester 1997) and pollen (Perveen et al. 2004) data. I accept Entelea as a genus endemic to New Zealand, although the monophyly of related genera needs to be investigated.

Hoheria A.Cunn. (7 spp., de Lange & Rolfe 2010; 6 spp., Breitwieser et al. 2012) and Plagianthus J.R. & G.Forst. (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

In phylogenetic studies based on ITS and cpDNA, Wagstaff et al. (2010) and Wagstaff & Tate (2011) found good support for the monophyly of Plagianthus and of Hoheria. I accept both Hoheria and Plagianthus as a genera endemic to New Zealand.

Montiaceae

Hectorella Hook.f. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Applequist et al. (2006, using matK and rbcL data) and Nyffeler & Eggli (2010, using trn K and matK data) found Hectorella to be sister to a clade comprising Lewisia, Montia and Claytonia, within a grouping recognised at family rank as Montiaceae (Nyffeler & Eggli 2010). Wagstaff & Hennion (2007, using matK, trnK and rbcL data; Fig. 14) also found this, and included Lyallia (from Kerguelen Island) in their study. Lyallia and Hectorella are sister taxa, as long suspected from morphological studies (Philipson & Skipworth 1961; Nyananyo & Heywood 1987). Both genera are monospecific, and differ in numbers of petals and stamens and their sexual systems (Hectorella is gynodioecious, whereas Lyallia is cosexual). Wagstaff & Hennion (2007) and Applequist et al. (2006) consider these differences sufficient to maintain the genera, contrary to the view of Nyananyo & Heywood (1987), who combined them in a single genus, Lyallia. Deciding between the alternatives may seem subjective, but I prefer to follow Nyananyo & Heywood (1987) because: (1) both flower merosity (Garnock-Jones 1993a) and sexual systems (Webb 1979; Garnock-Jones 1991b; Bayly & Kellow 2006) can be labile in other groups, and (2) maintaining two monospecific sister genera results in rank redundancy. The combination Lyallia caespitosa (Hook.f.) Nyananyo & Heywood is available.

Figure 14 Relationships of Hectorella (Montiaceae). Simplified parsimony and maximum likelihood tree derived from trnK and matK sequence data, parsimony bootstrap support values shown (from Wagstaff & Hennion 2007).

Myrtaceae

Lophomyrtus Burret (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Neomyrtus Burret (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Neomyrtus and Lophomyrtus were segregated from Myrtus by Burret (1941). Lucas et al. (2007) showed with moderate support from combined plastid and ITS sequence data that L. obcordata is more closely related to N. pedunculata than it is to L. bullata (Fig. 15). However, a phylogeny from ITS sequences alone had Neomyrtus sister to a monophyletic Lophomyrtus (Lucas et al. 2005). Monophyly of Lophomyrtus might be further supported by its 4-merous flowers and uniseriate ovules (Allan 1961), and the conflict between nuclear and chloroplast sequence analyses might indicate intergeneric chloroplast exchange. Thornhill et al. (2012) reported only minor differences in pollen morphology between Lophomyrtus and Neomyrtus. Further research is needed to test reciprocal monophyly of Lophomyrtus and Neomyrtus, and even if both are monophyletic, consideration should be given to combining them to reflect their close relationship.

Figure 15 Relationships of Neomyrtus and Lophomyrtus (Myrtaceae; Lucas et al. 2007: based on parsimony and Bayesian analyses of combined ITS, ETS, cpDNA, bootstrap values > 50 shown).

Nanodeaceae

Mida Endl. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Current published and unpublished knowledge on the status of Mida was summarised by de Lange (2011b). A proposed congeneric, M. fernandeziana of the Juan Fernández Islands (Der & Nickrent 2008), is now firmly established as belonging in Santalum (Bernardello et al. 2006; and see discussion by de Lange 2011b), which leaves Mida monospecific and endemic to New Zealand. Nickrent et al. (2010) place it with Nanodea of Patagonia and Tierra del Fuego in Nanodeaceae, a small clade of just two monospecific genera. Again, the issue of monospecific sister genera and redundant ranks arises, although Nickrent et al. (2010) describe major morphological differences between Mida and Nanodea.

Orchidaceae

There are 6–8 currently accepted (de Lange & Rolfe 2010; Breitwieser et al. 2012) endemic genera of orchids in New Zealand (but see de Lange et al. 2013); most are monospecific. Janes & Duretto (2010) note that whereas Australian botanists tend to adopt wider generic circumscriptions in orchids (they particularly refer to Pterostylidinae), New Zealand botanists have tended to adopt smaller segregate genera. For some of these (e.g. Ichthyostomum), detailed phylogenies are not yet available.

Aporostylis Rupp & Hatch (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

The phylogenetic tree of Clements et al. (2002, based on ITS sequence data) places A. bifolia in a large polytomy with many other Diurideae; further research is needed in order to clarify its relationships. The genus was accepted by www.e-monocot.org (accessed 27 September 2013). I accept Aporostylis as a genus endemic to New Zealand on current evidence.

Danhatchia Garay & Christenson (1 sp.; de Lange & Rolfe 2010, accepted but not as endemic by de Lange et al. 2013; Breitwieser et al. 2012)

Danhatchia australis was originally placed in Yoania (as Y. australis) by Hatch (1963), but he recognised its differences from other species of Yoania by proposing a new section, sect. Tarairea for it. Moore (in Moore & Edgar 1970) and Garay & Christenson (1995) elaborated the distinctive differences between this plant and Yoania, which remove it from Yoania (Tribe Cymbidieae) and instead support placement in subtribe Goodyerinae. In Goodyerinae, Garay & Christenson (1995) considered it to be related to Chamaegastrodia, but sufficiently different to be segregated as a new genus Danhatchia. The genus is widely accepted, including by www.e-monocot.org (accessed 27 September 2013). However, there is a recent online record of two observations of D. australis in Australia (http://plantnet.rbgsyd.nsw.gov.au/cgi-bin/NSWfl.pl?page=nswfl&lvl=sp&name=Danhatchia∼australis), so Danhatchia should no longer be considered endemic.

Earina Lindl. (3 spp.; de Lange & Rolfe 2010)

The current circumscription of Earina includes additional species in Fiji, Vanuatu, Samoa and New Caledonia (e.g. Lehnebach & Robertson 2004), so Earina should not be considered endemic.

Ichthyostomum D.L.Jones, M.A.Clem. & Molloy (1 sp.; de Lange & Rolfe 2010, but not de Lange et al. 2013 who include it under Bulbophyllum)

This monospecific genus was published by Clements & Jones (2002) to accommodate the species previously known as Bulbophyllum pygmaeum. They presented a diagnosis with the comment that several genera were being segregated ‘resulting from detailed morphological and molecular studies that will be published elsewhere.’ I have not found phylogenies that include B. pygmaeum and I note www.e-monocot.org (accessed 27 September 2013) and Felix & Guerra (2010) do not accept Ichthyostomum. Acceptance of Ichthyostomum before the presentation of supporting evidence is not justified.

Breitwieser et al. (2012) treated Ichthyostomum as non-endemic, apparently an error based on an erroneous record (Moore in Moore & Edgar 1970) of I. pygmaeum from Lord Howe Island (I. Schönberger pers. comm.), a record not supported by Rodd & Pickard (1983).

Molloybas D.L.Jones & M.A.Clem. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012) and Singularybas Molloy, D.L.Jones & M.A.Clem. (1 sp.; de Lange & Rolfe 2010, but not de Lange et al. 2013; Breitwieser et al. 2012)

Clements et al. (2002) combined ITS sequences and morphological data to derive relationships within Corybas s.l. They promoted numerous small clades to genus rank including Molloybas and Singularybas, both monospecific and endemic to New Zealand.

Molloybas (one species, endemic to New Zealand) and Nematoceras (seven species) were resolved as sister clades, together sister to Corysanthes (five species) (Fig. 16). Molloybas is morphologically distinctive, largely due to characters related to its non-photosynthetic nutrition. Clements et al. (2002) presented strong support for monophyly of Nematoceras (100% bootstrap; 10 ITS and 3 morphological characters) and Molloybas (24 autapomorphies, 8 of them morphological although several are likely to be functionally or developmentally linked). The sister relationship of Nematoceras and Molloybas received 84% bootstrap support (13 supporting characters, of which 6 morphological ones were listed).

Figure 16 Relationships among segregate genera (numbers of species shown) of the Corybas complex based on evidence from ITS sequence data and morphology (Clements et al. 2002; *, bootstrap value > 75%).

de Lange et al. (2013) returned most of these segregate genera (Anzybas, Nematoceras, Singularybas) to a more inclusive circumscription of Corybas. However, without explanation, they continued to accept Molloybas. Such a wider circumscription of Corybas is not monophyletic because some species treated as Corybas, for example, C. trilobus, are more closely related to Molloybas cryptanthus than they are to Corybas aconitiflorus.

With monophyly of both the wide circumscription of Corybas, including both Molloybas and Singularybas, and of the smaller segregate genera apparently well established, the issue resolves around the perceived advantages of large or small genera in this group, as discussed by Hopper (2009).

Singularybas (Corybas oblongus) has a large number of morphological autapomorphies (Clements et al. 2002) and its sister clade (six genera) also appears to enjoy moderate support (four synapomorphies, although none of them is unique in the wider tree). Its acceptance is similarly dependent on subjective preferences for narrow vs. wide generic circumscriptions in the Corybas complex.

Molloybas and Singularybas were treated as synonyms of Corybas by www.e-monocot.org (accessed 27 September 2013). I regard their acceptance at generic rank as equivocal on current evidence.

Waireia D.L.Jones, Molloy & M.A.Clem. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Clements et al. (2002) used ITS sequence data to demonstrate that Waireia stenopetala is not closely related to Lyperanthus, where it had previously been classified (as L. antarcticus). Instead, it is well supported (93% bootstrap support) as sister to another monospecific genus, Rimacola from Australia. The Rimacola–Waireia clade is placed among a largely unresolved group of other genera in a consensus tree (Clements et al. 2002, based on ITS sequence data). Although classifying them together in a genus would be more information-rich and would avoid rank redundancy, they are strikingly different in appearance and there seems little reason to introduce more names by challenging the status quo, especially while its wider relationships are not resolved. The genus was accepted by www.e-monocot.org (accessed 27 September 2013).

Winika M.A.Clem., D.L.Jones, & Molloy (1 sp.; de Lange & Rolfe 2010, but not de Lange et al. 2013; Breitwieser et al. 2012)

Winika is a monotypic genus segregated from Dendrobium (Clements et al. 1997). In publishing the new genus, Clements et al. (1997) described the plants in detail and compared them with Dendrobium sect. Kinetochilus from New Caledonia. However, Burke et al. (2008; Fig. 17) demonstrated that Winika cunninghamii is sister to the Australian Dendrobium clade and together these are sister to the Asian Dendrobium clade (which includes the type of the genus). Dendrobium sect. Kinetochilus was not included in those studies. If the Australian clade is separated from Dendrobium at generic rank, recognition of Winika might be appropriate, but there is little support from recent consensus trees (Burke et al. 2008; Adams 2011) for further subdivision of the Australian clade. Burke et al. (2008) and Adams (2011) discuss the options, and use the binomial Dendrobium cunninghamii in their papers. Conran et al. (2009) also followed a broad circumscription of Dendrobium (incl. Winika). The genus was treated as a synonym of Dendrobium by www.e-monocot.org (accessed 27 September 2013). In addition, the cultivar Dendrobium ‘Malcolm Campbell’ is a hybrid between D. kingianum of the Australian clade and D. cunninghamii (http://apps.rhs.org.uk/horticulturaldatabase/orchidregister/orchiddetails.asp?ID=145092). Dendrobium cunninghamii appears to represent a relatively early divergent lineage within Australasian Dendrobium s.l., and future research, which should include species from New Caledonia, might support it as an endemic New Zealand genus.

Figure 17 Summary of relationships within Dendrobium s.l. based on ITS sequence data after Burke et al. (2008). Bootstrap support shown.

Plantaginaceae

Heliohebe Garn.-Jones (6 spp.; de Lange & Rolfe 2010) and Leonohebe Heads (5 spp.; de Lange & Rolfe 2010)

Breitwieser et al. (2012) did not accept Heliohebe and Leonohebe as endemic genera because they treated the New Zealand genera Chionohebe, Hebe, Hebejeebie, Heliohebe, Leonohebe and Parahebe as part of Veronica (Garnock-Jones et al. 2007), whereas de Lange & Rolfe (2010) did accept Heliohebe and Leonohebe as endemic genera.

Both Heliohebe (Garnock-Jones 1993a,b; Wagstaff & Garnock-Jones 1998) and Leonohebe (Wagstaff et al. 2002; Bayly & Kellow 2006) are monophyletic, as circumscribed by Garnock-Jones (1993b) and Bayly & Kellow (2006), respectively, but other genera in the complex are not (Wagstaff et al. 2002; Albach & Meudt 2010; Fig. 18). Moreover, the paraphyly of Veronica with respect to the southern hebe complex is well established (Albach & Chase 2001; Wagstaff et al. 2002; Albach & Meudt 2010). Despite assertions that a broad circumscription of Veronica fails to consider other evidence than DNA sequences (Gardner 2007; de Lange & Rolfe 2010; de Lange et al. 2010), phylogenetic inclusion of the hebe complex within Veronica is consistent with evidence from morphology (Oliver 1944; Raven 1973; Garnock-Jones 1975, 1993a; Hong 1984; Kampny & Dengler 1997; Albach & Chase 2001; Hufford & McMahon 2004; Bayly & Kellow 2006; Muñoz-Centeno et al. 2006), cytology (Hair 1967, 1970; Albach & Greilhuber 2004; Briggs & Ehrendorfer 2006b; Albach et al. 2008), chemistry (Taskova et al. 2004, 2008, 2012; Albach et al. 2005; Pedersen et al. 2007; Jensen et al. 2008), while their tendency to hybridise between groups (Garnock-Jones & Lloyd 2004; Meudt & Bayly 2008) is evidence against multiple genera in New Zealand.

Figure 18 Relationships within southern hemisphere Veronica, showing phylogenetic positions of southern hemisphere segregate genera as accepted by de Lange & Rolfe (2010) and Mark (2012) (simplified from Albach & Meudt 2010, based on four cpDNA regions and ITS; posterior probabilities shown).

The paraphyly of Veronica and range of potential solutions, particularly with respect to Hebe and Leonohebe, were clearly discussed by Albach et al. (2004) and Bayly & Kellow (2006) and consequences for southern hemisphere plants were addressed nomenclaturally by Garnock-Jones et al. (2007) and Heenan (2012). Albach & Chase (2001), Albach et al. (2004), Garnock-Jones et al. (2007), Mabberley (2008), Albach (2008) and Albach & Meudt (2010) also presented arguments for including all the southern genera in a wide and monophyletic circumscription of Veronica. Briggs & Ehrendorfer (2006a), Jensen et al. (2008), Mabberley (2008), Meudt (2008), Taskova et al. (2008, 2012), Humphries & Linder (2009), Hughes et al. (2010), Pufal et al. (2010), Lee et al. (2011), Heenan (2012), Lord et al. (2013), and the Landcare Research herbarium and databases (see Breitwieser et al. 2012) have followed this wider circumscription. Nevertheless, other New Zealand botanists have explicitly rejected an inclusive monophyletic circumscription of Veronica (Gardner 2007; Thorsen 2007; Norton & Molloy 2009; de Lange & Rolfe 2010; de Lange 2011a; Mark 2012; de Lange et al. 2013; Wilson 2013). Their arguments have mostly focused on distinctiveness of the southern segregates, while Norton & Molloy (2009) and Mark (2012) also placed significance on their southern distribution. In my opinion, classifying them in Veronica among their northern hemisphere relatives provides a better indication of their relationships, evolution and biogeographic history, and avoids misleading inferences of relationships based on the earlier classification. While continued recognition of segregate genera (Gardner 2007; Thorsen 2007; Norton & Molloy 2009; de Lange & Rolfe 2010; de Lange 2011a; Mark 2012; de Lange et al. 2013; Wilson 2013) would preserve nomenclatural stability for New Zealand users, this is at the expense of the monophyly of Veronica, Parahebe, Chionohebe and Hebejeebie. Alternatively, if northern hemisphere segregate genera were introduced to avoid paraphyly of Veronica, this would be at the cost of introducing some that were undiagnosable with morphological characters and would result in changes to nearly 200 species names (Albach et al. 2004; Garnock-Jones et al. 2007; Garnock-Jones & Albach 2007). The arguments for inclusion of the hebe complex in Veronica are the same as those for inclusion of Neopaxia within Montia, Iti within Cardamine, Macropiper within Piper and Exarrhena within Myosotis, which have been widely accepted by New Zealand botanists.

Poaceae

Anemanthele Veldkamp (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Anemanthele appears to be either nested within or sister to Austrostipa, according to the molecular phylogenetic studies of Jacobs et al. (2000, 2007) and Romaschenko et al. (2012: ITS and nine plastid regions). Romaschenko et al. (2012) found conflicting results on the relationship between Anemanthele and Austrostipa. First, based on cpDNA, they found Anemanthele to be sister to Austrostipa. Second, using ITS data, Anemanthele was sister to one of two Austrostipa clades. If the non-monophyly of Austrostipa found in the ITS tree is supported by further studies, two taxonomic solutions would involve including Anemanthele within either Austrostipa or a part of Austrostipa, removing it from New Zealand's endemic genera. A three-genus solution would retain a monotypic endemic Anemanthele, which would reflect its morphological differences (Entwisle & Weston 2005; Jacobs et al. 2007). I accept Anemanthele only provisionally as a genus endemic to New Zealand.

Austroderia N.P.Barker & H.P.Linder. (5 spp.; de Lange 2011a; Breitwieser et al. 2012)

The genus was proposed by Linder et al. (2010) to accommodate a clade of New Zealand species previously treated under Cortaderia. The published phylogeny of Barker et al. (2007a) is complex because of conflicts between the signals from nuclear and cpDNA (Fig. 19A). However, it is clear that Austroderia is a clade and is more closely related to Plinthanthesis than it is to Cortaderia. Linder et al. (2010) have recognised clades at generic rank that are all diagnosable and well supported (including Austroderia), rather than an alternative of a large and highly varied genus Danthonia. I accept Austroderia as a genus endemic to New Zealand.

Figure 19 Relationships in New Zealand Poaceae. A, Simplified phylogeny showing relationship of Austroderia and Cortaderia, following Linder et al. (2010 Fig. 2; based on 8200 bp of cpDNA from trnL intron, trnL–F spacer, rpl16 spacer, atpB–rbcL spacer, ndhF gene, matK gene and flanking regions and rbcL gene, and 1800 bp of nuclear DNA from the ITS region and 26s rDNA; posterior probabilities shown). B, Relationships of Zotovia, Ehrharta and Microlaena, after Verboom et al. (2003, based on combined ITS and trnL–F sequences and morphology; bootstrap support values shown).

Pyrrhanthera Zotov (1 sp.; de Lange & Rolfe 2010, but not de Lange 2011a, de Lange et al. 2013; Breitwieser et al. 2012)

Barker et al. (2007a) analysed data from ITS and three chloroplast regions (rpoC2, trnL intron, rbcL gene) and showed the single species of Pyrrhanthera is nested within Rytidosperma. They treated it as R. exiguum (Kirk) H.P.Linder, a classification that I accept.

Simplicia Kirk (2 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

There appears to be no current doubt that Simplicia (two spp.) is acceptable as a genus and endemic to New Zealand. Gillespie et al. (2009) cited a dissertation by Döring (2009, not seen) who placed it in Poinae and showed Simplicia is not nested within Poa. I accept Simplicia as a genus endemic to New Zealand.

Stenostachys Turcz. (3 spp., de Lange & Rolfe 2010; 4 spp., Breitwieser et al. 2012)

Stenostachys is a small New Zealand endemic genus with an allopolyploid origin involving Australopyrum and Hordeum, according to the molecular studies of Petersen et al. (2011), who support its recognition as a genus based on analysis of two single-copy nuclear genes, four chloroplast genes and one mitochondrial gene. They support the transfer of Elymus enysii to Stenostachys (Barkworth & Jacobs 2011) additional to the three species accepted by Edgar & Connor (2000). I accept Stenostachys as a genus endemic to New Zealand.

Zotovia Edgar & Connor (3 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Verboom et al. (2003) presented a phylogeny based on combined cpDNA, ITS1 and morphology data that showed Zotovia is nested within Ehrharta as currently circumscribed (Fig. 19B). Support for this relationship was weak, and the sister relationship of Z. colensoi and M. avenacea was supported only by the ITS data. If these relationships are supported by additional research, both Zotovia and Ehrharta could acquire monophyletic circumscriptions by the transfer of E. avenacea, Microlaena avenacea (these two names have different types and are not considered conspecific) and M. tasmanica to Zotovia (Verboom et al. 2003; Kellogg 2009). The genus would not then be endemic. Alternatively, those three species could be classified as a sister genus to Zotovia, which latter would then still be endemic. I regard generic status of Zotovia as equivocal on current evidence.

Primulaceae

Elingamita G.T.S.Baylis (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Baylis (1951) compared Elingamita johnsonii with Myrsine and other genera, noting that although it matches Myrsine in its uniseriate ovules, it differs in its paniculate inflorescences, short tubular corolla, well-developed filaments and punctiform stigma. Although Stöckler (2001: ITS, ndhF and two AFLP-derived markers) suggested that E. johnsonii is nested within New Zealand Myrsine, ITS sequences of E. johnsonii independently obtained by P.B. Heenan (pers. comm.) and C.E.C. Gemmill & S. Pratt (pers. comm.) did not support a close relationship of Elingamita within Myrsine. Current research is focused on placing Elingamita in relation to Myrsine and other genera in Primulaceae and testing monophyly of its relatives (C.E.C. Gemmill pers. comm.). I accept Elingamita provisionally as a genus endemic to New Zealand.

Proteaceae

Knightia R.Br. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Recent circumscriptions of Knightia have included one New Zealand and two New Caledonian species (e.g. Virot 1968; Jaffré et al. 2001). However, Weston & Barker (2006) treated the New Caledonian species as Eucarpha Spach, based on a phylogenetic supertree that combined several molecular studies. In that tree, Knightia grouped with Roupala, Orites and Neorites, whereas Eucarpha grouped with Floydia and Darlingia. Sauquet et al. (2009) presented a phylogeny of Proteaceae, based on nuclear ITS and chloroplast atpB, atpB–rbcL, rbcL, matK, rpl16 intron, trnL intron and trnL–F, in which Knightia was sister to a clade comprising nine genera, one of them Eucarpha (Fig. 20A). Their study indicated a Paleogene age for Knightia. I accept Knightia as a genus endemic to New Zealand.

Figure 20 Relationships of New Zealand Proteaceae. A, Relationships of Knightia, simplified from Sauquet et al. (2009, Bayesian tree; data from nuclear ITS and chloroplast atpB, atpb–rbcL, rbcL, matK, rpl16 intron, trnL intron and trnL–F). B, Relationships of Toronia, simplified from Sauquet et al. (2009, Bayesian tree; data from nuclear ITS and chloroplast atpB, atpb–rbcL, rbcL, matK, rpl16 intron, trnL intron and trnL–F).

Plants from New Caledonia assigned to Eucarpha differ from K. excelsa in being shrubs or small trees, having smaller flowers and enlarged showy bracts that subtend each floral pair, and woody follicles that dehisce partly along the midrib as well as fully at the margins (Virot 1968). They differ also in their pollen: K. deplanchei² grains differ from those of K. excelsa by having convex sides, circular and non-protruding pores, weakly developed vestibule and thinner exine with smaller and more scattered bacculae (P.J. Garnock-Jones, unpubl. data).

Toronia L.A.S.Johnson & B.Briggs (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Toronia, a monospecific genus from New Zealand, was previously included in Persoonia, as P. toru (Allan 1961). Weston & Barker (2006), Barker et al. (2007b), Hoot & Douglas (1998) and Sauquet et al. (2009) accepted Toronia and included it in Persoonioideae. Sauquet et al. (2009; Fig. 20B), based on ITS and seven cpDNA regions, showed Toronia to be sister to Persoonia (100 spp.), Acidonia (1) and Garnieria (1). However, only a few samples of Persoonia have been included in those studies, and Weston & Barker (2006) referred to unpublished results of Weston & Porter that suggest Persoonia is not monophyletic. Acceptance of Toronia should thus be provisional until those studies are completed.

Barker et al. (2007b) calculated the divergence of Placospermum and Toronia within a range that includes the 80 million year old separation of the New Zealand land mass from Gondwana, but Sauquet et al. (2009) indicate divergence of Toronia is of Miocene age.

Strasburgeriaceae

Ixerba A.Cunn. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

The relationships of Ixerba have been a long-standing enigma (Cameron 2003). Recently, DNA sequence data (Sosa & Chase 2003; Qiu et al. 2010; Fig. 21), flower morphology (Matthews & Endress 2006) and wood anatomy (Patel 1973; Cameron 2003; Carlquist 2007) all support the sister relationship of Ixerba brexioides and Strasburgeria robusta. Although they share a base chromosome number of x = 25, Ixerba (2n = 50) and Strasburgeria (2n = 500) have very different chromosome numbers (Oginuma et al. 2006). Maintaining these as two monotypic families or genera appears to be a subjective question. Cameron (2003) listed differences and similarities and argued that maintaining two families was justified; this has been maintained by others (Matthews & Endress 2006; Dickison 2007; Schneider 2007; Endress & Matthews 2012). New Zealand botanists (de Lange & Rolfe 2010; Breitwieser et al. 2012) have followed the Angiosperm Phylogeny Group (APG III 2009) and Qiu et al. (2010), including Ixerba in Strasburgeriaceae, a classification that increases information content and decreases rank redundancy. In this case, combining the two species under a single genus would neither increase information content nor decrease rank redundancy and I accept Ixerba as a genus endemic to New Zealand.

Figure 21 Relationships of Ixerba, simplified from Qiu et al. (2010, data from four mitochondrial gene sequences, atp1, matR, nad5, and rps3; bootstrap support values shown).

Stylidiaceae

Oreostylidium Berggr. (1 sp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Oreostylidium subulatum is nested within the Australian genus Stylidium (Wagstaff & Wege 2002; Fig. 22). The name Stylidium subulatum Hook.f. is available and preferred, for example, by Wege (2012). The New Zealand species appears to be distinctive because it lacks the irritable column and zygomorphic corolla of Stylidium in Australia, but the phylogeny presented by Wagstaff & Wege (2002) would suggest these are secondary losses and not unique in the genus (Wege 2012).

Figure 22 Relationships of Stylidium subulatum (as Oreostylidium subulatum), after Wagstaff & Wege (2002, based on ITS and rbcL sequence data; bootstrap support values shown).

Winteraceae

Pseudowintera Dandy (4 spp.; de Lange & Rolfe 2010; Breitwieser et al. 2012)

Doust & Drinnan (2004) used ITS and trnL sequence data and morphology to study the relationships and monophyly of Tasmannia (Winteraceae). They found strong support for a monophyletic Pseudowintera (two of the four species were sampled), strongly supported as sister to a clade of Belliolum, Bubbia, Exospermum and Zygogynum. I accept Pseudowintera as a genus endemic to New Zealand.

Discussion

This re-evaluation of the taxonomic and distribution status of up to 59 seed plant genera that are treated as endemic in New Zealand has reduced the number to between 28 and 44. Fifteen genera, all angiosperms, should no longer be accepted as endemic or should not be accepted at generic rank. For these, either the monophyly of a related assemblage has been falsified (e.g. Kirkianella and Embergeria, Heliohebe and Leonohebe, Oreostylidium), or they have been shown to occur outside New Zealand (e.g. Danhatchia, Earina, Stilbocarpa). The status of a further 16 genera is equivocal and these require further reconsideration of their relationships and status. Thus, about half of the currently accepted endemic genera in New Zealand are taxonomically rejected or equivocal or they are not endemic. Even in some pairs of genera for which reciprocal monophyly has not been conclusively falsified (e.g. Lophomyrtus and Neomyrtus, Teucridium and Teucrium), currently available research indicates uncertainty, and a few of them could more informatively be classified in the same genus as close relatives in New Zealand and overseas.

This major reduction in the number of endemic genera recognised in New Zealand is a logical taxonomic outcome of recent molecular studies of relationships (Mitchell & Heenan 2000; Wagstaff & Wege 2002; Wagstaff & Breitwieser 2004) and a corollary of recent biogeographic research (Winkworth et al. 2002b) in the New Zealand flora. There has been widespread recognition in the last 15 years that the flora is mostly of Oligocene or more recent origin in New Zealand and very few groups are Mesozoic relicts. The very recent uplift of the New Zealand mountains (Heenan & McGlone 2012), where most of the plant species diversity is found, correlates well with the young age of the flora. In an analysis of Tasmanian endemic genera, Rozefelds (2008) divided them into paleoendemics and neoendemics, and showed that members of these two groupings tended to have different distributions. In New Zealand, a few endemic genera with a Paleogene (65–23 Ma) or older advent (e.g. Ixerba, Knightia, Rhabdothamnus) are found in northern and lowland sites, whereas some of the likely neoendemics (e.g. Notothlaspi, Raoulia) are southern and alpine. However, we must await better estimates of dating before such an analysis can be completed for New Zealand endemic genera.

McGlone et al. (2001) calculated that c. 10% of vascular plant genera (including ferns) in New Zealand are endemic, compared with 15% in Hawaii and 13.5% in New Caledonia. Here, I provide an estimate of 7–10.5% for the seed plants. My upper bound is higher than the estimate of McGlone et al. (2001) because their figure included ferns, a group with low endemism. Such estimates are highly susceptible to taxonomic concepts and methodologies. Floras with a long history of isolation from other landmasses would be expected to have more endemism overall and more endemism at higher ranks than do young floras. However, if phenetic criteria are widely used in a young and rapidly diversifying island flora, high levels of inferred endemism might be expected, and this might be misinterpreted as a greater age for the flora. Where generic classification is partly or wholly based on phenetic considerations, it is likely that generic endemism in island floras will be overestimated.

Some families appear to have greater numbers of endemic genera than their size, or their size in New Zealand, would suggest. Thus Loranthaceae, with just 7 species in New Zealand (Breitwieser et al. 2012), has 3 or 4 endemic genera, compared with 84 genera for 950 species world-wide (Mabberley 2008). It may be that in some families, such as Loranthaceae and Balanophoraceae, our few species are geographic outliers and their adaptations to local conditions have led to their segregation at genus rank. In Brassicaceae, phylogenetic systematics has led to a reduction in the number of recognised endemic genera from four (Garnock-Jones 1991a) to the current one (Heenan et al. 2002; but see Heenan et al. 2012).

The previous recognition of 50–60 endemic genera is not just an artefact of an earlier view that the New Zealand flora is ancient and has evolved here in long isolation. Many of the genera rejected here display considerable morphological and ecological divergence, which encouraged their separation in a pre-phylogenetic era. However, recognising their sister relationships through phylogenetic classification enables a more realistic appraisal of their evolutionary history. For example, treating Haastia as an endemic genus gives no hint that these distinctive vegetable sheep are nested within Brachyglottis, whereas including them in Brachyglottis makes it immediately clear that this is an example of the evolution of a cushion-plant clade within a genus of shrubby plants, and is an instructive parallel to the distinctive cushion-plant clades within Raoulia, Veronica and Myosotis.

Sonchus is a good example of a widespread and otherwise coherent genus that has diverged morphologically on islands or in extreme habitats (Kim et al. 2007). Small distinctive locally endemic clades are nested within Sonchus in New Zealand (previously treated as Embergeria and Kirkianella), Juan Fernández Islands (treated as Dendroseris until Mejías & Kim 2012) and Australia (Actites), while a diverse generic complex in Macaronesia (Babcockia, Chrysoprenanthes, Lactucosonchus, Sventenia, Taeckholmia) appears to be sister to Sonchus s.l.

However, there are probably other genera that are not currently considered to be endemic, but for which changes to current circumscriptions may be needed to make them monophyletic. The approach adopted here has taken the currently recognised endemic genera and subjected them to consistent criteria based upon the primacy of clade recognition and analyses based on molecular phylogenetics. It is important to note that, by starting with such a list, a reduction was probably likely. Critical assessment of the status of all genera in the flora might, however, lead to the recognition of additional endemic genera, increasing the total for the country, as in the following three examples. First, the current circumscription of Pseudopanax (Araliaceae) may be paraphyletic (Nicolas & Plunkett 2009) and the authors resolve this by accepting the segregate genus Neopanax, which was considered endemic when Allan (1961) first named it. Additionally, in Araliaceae, Mitchell et al. (2012) showed the current circumscription of Raukaua is paraphyletic and Australian and South American species should be excluded. Potentially, an emended circumscription of Raukaua might be endemic to New Zealand, but this depends on whether species of Cheirodendron from Hawaii and the Marquesas Islands should be included.

Second, Ischnocarpus and Pachycladon (Brassiaceae), formerly endemic genera, are now combined (Heenan et al. 2002) with the non-endemic Cheesemania in a wider circumscription of Pachycladon. Heenan et al. (2012) showed that including the Tasmanian P. radicata in the circumscription of Pachycladon makes the genus non-monophyletic. This justifies taxonomic exclusion of P. radicata, and that would make Pachycladon endemic to New Zealand. Additionally, except for P. wallii, the three former genera form morphologically distinctive endemic clades that have weak molecular support (the former C. wallii has morphological and molecular similarities to P. novae-zelandiae and P. crenatus, and shares their southern distribution—Garnock-Jones 1991b; Heenan et al. 2002). However, fertile hybrids between these clades have been produced (Heenan 1999b) and retention of a single endemic genus might be preferable.

Third, the current paraphyletic circumscription of Olearia (Asteraceae) will need to be revised. If Celmisia and Pachystegia are to be retained, Olearia will need to be divided into several clades (Cross et al. 2002). Some of these, for example, Shawia, might be endemic New Zealand genera.

Additionally, it is not immediately clear how to list genera that have lost their endemic status through recent migration (e.g. Ileostylus, Mills 2010) or to account for the possibility of a genus becoming endemic through extinction of a foreign population.

In any case, it is pertinent to question the significance of genera in analyses of the flora such as those of McGlone et al. (2001), Webb et al. (1999) and Lee et al. (2011). The genus is just one of many ranks in the classification hierarchy, and some of its significance derives from the binomial system, wherein the genus name forms part of the species name. Choosing to analyse the flora at the rank of genus or of family is simply using the rank as a proxy for clades that might be equivalent in size and significance. However, one should not assume that genera, even if they are monophyletic, are in any way necessarily equivalent. Some kinds of analyses should focus instead on clades of any rank. For example, treating current genera as data points in analyses could over-represent some groups, such as the Raoulia complex (currently including four endemic genera) and under-represent others (clades within Pachycladon). Additionally, excluding from such analyses clades that originated and diversified in New Zealand, but which through dispersal are not endemic (e.g. Veronica sect. Hebe or the Aciphylla–Anisotome complex) could also be misleading. An analysis of endemism that compared the most inclusive endemic clades in terms of their size, fossil or molecular age of origin in New Zealand, degree of morphological and physiological divergence, internal disparity of morphology, physiology, habitat requirements, or sexual systems would be revealing.

Acknowledgements

Several people helped with discussions and information: Eric Knox (Campanulaceae); Peter de Lange (orchids and grasses); Steve Pratt, Chrissen Gemmill, Peter Heenan and Karin Stöckler (Elingamita); and Angela Moles. I thank Chrissen Gemmill, Heidi Meudt, and two anonymous referees who read and commented on an earlier draft.

Notes

1. The epithet hortensia is a noun in apposition (www.ipni.org/index.html). Hortensia is a synonym of Hydrangea, of which some species have dense corymbs of blue flowers like those of Myosotidium. The epithet is not to be altered to hortensium as by Heenan & Schönberger (2009) and de Lange et al. (2010, 2013).

2. It appears that although the name Eucarpha is legitimate, combinations for the two species in New Caledonia have not been made. Therefore, it is necessary to refer to them as Knightia until new combinations (Weston & Garnock-Jones in prep.) are published.