Quantifying errors and omissions in alien species lists : The introduction status of Melaleuca species in South Africa as a case study

Introduced species lists provide essential background information for biological invasions research and management. The compilation of these lists is, however, prone to a variety of errors. We highlight the frequency and consequences of such errors using introduced Melaleuca (sensu lato, including Callistemon) species in South Africa as a case study. We examined 111 herbarium specimens from South Africa and noted the categories and sub-categories of errors that occurred in identification. We also used information from herbarium specimens and distribution data collected in the field to determine whether a species was introduced, naturalized and invasive. We found that 72% of the specimens were not named correctly. These were due to human error (70%) (misidentification, and improved identifications) and species identification problems (30%) (synonyms arising from inclusion of Callistemon, and unresolved taxonomy). At least 36 Melaleuca species have been introduced to South Africa, and field observations indicate that ten of these have naturalized, including five that are invasive. While most of the errors likely have negligible impact on management, we highlight one case where incorrect identification lead to an inappropriate management approach and some instances of errors in published lists. Invasive species lists need to be carefully reviewed to minimise errors, and herbarium specimens supported by DNA identification are required where identification using morphological features is particularly challenging. Copyright Llewellyn E.O. Jacobs et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. NeoBiota 32: 89–105 (2017) doi: 10.3897/neobiota.32.9842 http://neobiota.pensoft.net RESEARCH ARTICLE Advancing research on alien species and biological invasions A peer-reviewed open-access journal


Introduction
Species lists form the basis for much of the current research on biological invasions (e.g. the Global Naturalized Alien Flora Database of van Kleunen et al. 2015).Such lists are also essential for guiding legislation, as input to decision making and risk assessment, and in the formulation of management policies and strategies (McGeoch et al. 2012;Latombe et al. 2016).Because resources required to address the threat of invasive species are limited, objective categorization of species is required to prioritize resource allocation according to species, areas and introduction pathways (McNeely et al. 2001, Nel et al. 2003, Wilson et al. 2013).Accurate lists of alien species, with data on their introduction status, are thus crucial resources, not just for the regions for which they are compiled, but also globally (Wilson et al. 2011).But, as with any information derived from a variety of sources, the compilation of lists is prone to a number of errors which can then be perpetuated in various ways (McGeoch et al. 2012).To address these concerns, it is thus essential that the error rates in species lists are assessed and ways to detect them are identified.Pyšek et al. (2002) include alien taxa and their status in a flora of the Czech Republic.This well-compiled list lends itself to comparison with other regions and is an important example allowing for determinants and patterns of plant invasions at a global scale to be studied (Pyšek et al. 2004).Such lists are the essential building blocks on which assessments of the status of invasions in a country should be built (Latombe et al. 2016).By comparison, even though South Africa has a reasonably well-funded national programme for controlling invasive species, especially plants, research on lesser known invasive groups has only recently been given special attention (Wilson et al. 2013), and there is no comprehensive list of introduced and invasive species yet (Faulkner et al. 2015).A list of regulated invasive plant species was published in 2014 and this forms the basis for management plans and regulation (Department of Environmental Affairs 2014).However, this regulatory list is incomplete and contains several errors (per.obs.).Moreover, more species will need to be added as surveillance progresses, as more species demonstrate invasiveness, impacts are evaluated, and as errors in the list are discovered (Rouget et al. 2016).
For plants, herbaria are indispensable resources and reference sources for much botanical research which requires reliable species identifications, including the compilation of introduced species lists (Glen 2002).Funding for taxonomy and the upkeep of herbaria is declining worldwide (Smith et al. 2008, Guerin 2013, Pyšek et al. 2013) and is a concern that can be compounded because expertise for alien species is less likely to exist in any particular country.Herbarium specimens, upon which comprehensive lists are ideally based (Pyšek et al. 2013), require curation as taxa are revised or new information becomes available, e.g. from molecular and other studies (e.g., Le Roux et al. 2010).Many alien taxa are underrepresented, remain unidentified for considerable periods of time, or are misidentified in herbaria (Pyšek et al. 2013).In this paper we explore the scale of this problem using taxa in the genus Melaleuca (sensu Craven (2006) and Brophy et al. (2013)) in South Africa as a case study.
The genus Melaleuca has not been distributed around the world as extensively as some other tree groups (e.g.Eucalyptus, a sister genus in the Myrtaceae) (Rejmánek and Richardson 2011).However, seven species are listed as invasive in the USA and South Africa (Rejmánek and Richardson 2013), including one of the world's poster-child plant invaders, Melaleuca quinquenervia, which has invaded large areas and caused major damage in the Everglades region in Florida (Richardson and Rejmánek 2011).The genus has about 290 species consisting of shrubs and trees, a number of which are planted in many parts of the world, largely as ornamentals, but also for timber, honey, bark and plant extracts (Brophy et al. 2013).Widespread cultivation of Melaleuca species is relatively recent, especially when compared to other genera in the Myrtaceae such as Eucalyptus, and records of naturalization and invasions in South Africa (Jacobs et al. 2014(Jacobs et al. , 2015) ) and other parts of the world (Rejmánek and Richardson 2013) are comparably recent.Several species are recorded as weedy within Australia (Randall 2007), perhaps indicating that these (mostly) fire-adapted species could pose a risk to areas with similar fire-prone areas, such as the Cape Floristic Region of South Africa which has been invaded by many other woody plants from Australia (Wilson et al. 2014b).
In 2009, the discovery of several naturalised populations of Melaleuca species in South Africa prompted an evaluation of the introduction status for the entire group in the country (Wilson et al. 2013).Taxa such as Melaleuca armillaris subsp.armillaris, M. viminalis subsp.viminalis and M. citrina have been widely planted in South Africa and also warranted further study.This also provided an opportunity to reassess the accuracy of current published lists.
Here, we compile a list of Melaleuca species recorded as present in South Africa and determine the invasive status of each species.We use herbarium specimens to do this, while also noting the extent to which they are accurately identified and the types of errors which occur.We discuss consequences of errors and omissions and make recommendations on how these could be avoided and addressed.

Taxonomy
Generic limits in the tribe Melaleuceae have been the subject of much recent study (Brown et al. 2000, Wilson et al. 2005, Craven 2006, Edwards et al. 2010, Udovicic and Spencer 2012, Craven et al. 2014).We follow Craven (2006), Edwards et al. (2010) and Brophy et al. (2013) in adopting a broad concept of Melaleuca, i.e. including Callistemon.The further expansion of the genus Melaleuca to include Beaufortia, Calothamnus, Conothamnus, Eremaea, Lamarchea, Petraeomyrtus, Phymatocarpus, and Regelia (Craven et al. 2014), has not yet been fully evaluated by the Australian taxonomic community, and these taxa are excluded from consideration for this study.Many Melaleuca species (especially those formerly recognised as Callistemon) are morphologically similar which makes them difficult to identify using morphological features.Several cultivars have been developed for some Melaleuca species in the Callistemon group and difficulty in identifying such specimens in South Africa is perhaps due to horticultural selection and the existence of hybrid and both sexual and apomictic species within the group (Craven 2009, Brophy et al. 2013).

Review of herbarium specimens and error classification
Herbarium specimens from the Compton herbarium (NBG) were examined to check whether specimen identifications were correct, and to provide accurate identifications where necessary.To do this, we used the taxonomic literature to compare morphological characters on the specimens with descriptions and taxonomic keys (in particular Craven andLepschi 1999 andBrophy et al. 2013).Photographs and high-resolution scans of the specimens were taken for verification and future reference.Specimen identifications were checked against referenced herbarium specimens housed at the Australian National Herbarium (CANB; herbarium codes follow Thiers (2016)).The identifications of all specimens were subsequently confirmed by a taxonomic authority for Melaleuca (B.J. Lepschi).
Herbarium specimens were examined in 2013; any specimens accessioned or reidentified after this date were not included in the analysis.McGeoch et al. (2012) proposed an uncertainty classification that separates epistemic and linguistic errors into ten sub-categories.In this study we focussed on two of these sub-categories-human error and species identification.In keeping with terminology from McGeoch et al. (2012), we define the word "error" to be inclusive of actual and potential errors.For example, although a species name on a specimen was not currently accepted but no obvious mistake in listing arose from this yet, it was still recorded.As per McGeoch et al. (2012) scheme the human errors we discovered in this study were: misidentifications, and improved resolution of the identification (e.g.Melaleuca sp.identified as M. parvistaminea, or M. armillaris as M. armillaris subsp.armillaris).The only species identification error was unresolved taxonomy.A description of the different errors and how they were determined is shown in Table 1, as well as the frequency and relative proportions of the errors.Because the inclusion of Callistemon in an expanded Melaleuca is still under debate, synonyms where the genus name Callistemon changed to Melaleuca were placed under the species identification error type (instead of human error as per McGeoch et al. 2012 treatment).No synonyms outside of this situation were found and therefore synonyms relating to human error were absent from our dataset.
We also looked to see if there were any historical trends in the errors by comparing the years when taxa with particular errors were collected to the years when taxa with no errors were collected using Mann-Whitney U tests in R.
Table 1.Result of analysis of confirmed herbarium records (n=111), indicating the breakdown of correctly identified specimens with various error types.For full details see Suppl.material 1: Appendix.The errors identified here are error type 1 (i.e.human error, indicated as HE) and type 3 (i.e.species identification indicated as SI) as per McGeoch et al. (2012); synonyms are included in type 3 here (see Methods).The table only includes samples from the Compton Herbarium, Kirstenbosch (NBG).

Correctly identified
The identification on the herbarium specimen was the same as determined by an expert in the group (the author: BL)

List compilation
Once correct identification for all specimens had been confirmed, we used these specimens as the source for compiling a list of species present in South Africa.We also used a list of cultivated plants based on herbarium records in southern Africa (Glen 2002), and a list of forestry trees and their uses in South Africa (Poynton 2009).The minimum residence time in South Africa was determined from the date on the oldest herbarium specimen for each species.Naturalized populations were reported by a variety of conservation agencies, with the reports collated by the South African National Biodiversity Institute's Invasive Species Programme and through the Southern African Plant Invaders Atlas (Henderson et al. 2007;Wilson et al. 2013).
We collected height data as an estimate of age, presence/absence of reproductive structures and GPS coordinates for each plant.Using these data we were able to determine whether a species is sustaining itself, whether it is reproducing and/or spreading, hereby indicating the status of each species as introduced, naturalized or invasive according to the subcategories proposed by Blackburn et al. (2011).

Review of herbarium specimens
A summary of the errors found is in Table 1 with details of each specimen that required a name change in Suppl.material 1: Appendix.Examples of the types of errors on are shown in Figure 1.Of the 111 specimens examined, only 31 specimens carried a currently accepted name (excluding current names for specimens that were incorrectly identified).Misidentifications made up the largest proportion of errors, while poorly resolved taxonomy was the reason for two specimens being incorrectly named.All synonyms required at least the genus name to be changed.
There was no significant effect of date of collection on whether an error was noted, or on particular errors types (dates of collection varied between 1907 and 2013).

List compilation
Our analysis of herbarium specimens and the lists in Glen (2002), also based on herbarium collections, is summarised in Table 2 (no additional species were found in Poynton's (2009) list), with species that did not have confirmed herbarium records discussed in Table 3. Thirty-six species are confirmed present in South Africa, of which ten species are naturalized -five of these are invasive (Fig. 2; Table 2).Five naturalized species were categorised as C3 according to Blackburn et al. (2011), indicating that individuals were surviving, reproducing and populations were self-sustaining, but less than 100 m from planting sites (Richardson et al. 2000;Wilson et al. 2014).Melaleuca linearis var.linearis, M. hypericifolia, M. rugulosa and M. viminalis subsp.viminalis are invasive, surviving and reproducing a significant distance from the site of original introduction, but not over a wide extent (D2).Melaleuca parvistaminea is invading several sites (E) near the towns of Tulbagh and Wolseley in the Western Cape province (Fig. 2c).There are a few separate invasive populations spread over ~10,000 ha, with a total of around 30 000 plants (Fig. 3; Jacobs et al. ( 2014)).

Discussion and conclusions
There are a number of ways that errors can be generated during the compilation of species lists (McGeoch et al. 2012), but here we show the challenges that exist at a fundamental stage of the listing process.Importantly, since only a subset of herbaria were analysed in detail, there could be additional errors (and in fact additional species) present in South Africa.The high proportion of misidentifications (Table 1) is concerning, indicating the difficulties encountered when dealing with novel species and highlighting the need for expertise on specific non-indigenous taxa.Synonymy, however, does not necessarily imply    human error, but rather that the use of an outdated or otherwise superseded taxonomy can lead to errors in interpretation, or incorrect estimates of numbers by subsequent users (McGeoch et al. 2012).In this study however, synonymy arose rather from differing perceptions of Callistemon, than from human error.Although the effect of synonymy is potentially large (McGeoch et al. 2012), the checking of synonymies is commonly practised.However, a rudimentary training in taxonomic principles is necessary for any practitioner dealing with scientific names.It is of concern that the inclusion of Callistemon in an expanded Melaleuca is still under debate.All synonym issues found in our study at least  Melaleuca nervosa (Lindl.)Cheel Unknown Gibbs (1998) One tree recorded at Damara Farm near Malmesbury.Several Acacia species trials were also carried out at this site required the genus name to be updated.Lists not taking this into account could generate errors of a greater magnitude than errors relating only to the specific epithet.Lists therefore require the application of taxonomic expertise on taxa not native to a particular region (Pyšek et al. 2013).The knowledge generated from these lists form the basis for informing end users, (e.g.quarantine officials, conservation agencies) that perform crucial functions in stemming the tide of biological invasions and informing future research (e.g.identifying biological control agents) requiring accurate species identifications.Herbaria have often served as barometers for new and rediscovered alien plant species.They also provide a reference source for research or conservation initiatives that require accurate species names.The ongoing decline of resources being allocated to the maintenance of herbaria worldwide will adversely affect many research fields including invasion biology (Guerin 2013, Pyšek et al. 2013).We strongly believe that part of the funding for invasive species management needs to be allocated to the maintenance and functioning of herbaria and other collections as they are an essential resource for the work (this has begun to be supported in South Africa but further sustained resources need to be devoted to this).The same could be suggested for other fields of botanical research.
While genetic verification of species identifications is proving to be a reliable means of verifying a species, classical taxonomy still remains crucial to the identification of new species to a region (Pyšek et al. 2013).In the absence of molecular data suitable for use in species identification, identifications based on morphology are usually adequate (Pyšek et al. 2013).For these reasons, and an uncertain taxonomy in some cases, we found morphological identification based on published descriptions and keys the best approach to reviewing herbarium specimens of Melaleuca.Because suitable molecular data is often lacking, we recommend that DNA barcoding efforts should prioritise potentially invasive genera, so that species can be accurately identified in regions where expertise on that group is likely absent.Species identification issues due to uncertain or unresolved taxonomy can be avoided by continued taxonomic research (Edwards et al. 2010).This research will likely be conducted in the country of origin and therefore cross-border communication and collaboration between taxonomists are essential (Smith et al. 2008, Pyšek et al. 2013).Errors could be avoided by either collaborating with researchers from regions where alien species are native, thus tying into a strategic response of the Global Invasive Alien Species Strategy (McNeely 2001) or by investing in local taxonomic expertise on key alien groups.There are several ways in which these groups could be identified based on known patterns of invasion.Minimum residence time, invasiveness in other regions and weedy species are data obtainable from herbarium specimens and could thus be used to identify these groups.
Identification errors noted in this study have had direct implications.Melaleuca parvistaminea was initially misidentified in 2011 as the morphologically similar M. ericifolia.Melaleuca parvistaminea was only formally described in 1984 and collections prior to this were treated within the broad concept for M. ericifolia.Some M. armillaris subsp.armillaris specimens were also misidentified as M. ericifolia (e.g.NBG0269364).Melaleuca ericifolia is regarded as being predominantly clonal rather than reseeding.This affected management actions, through unforeseen profuse recruitment via seed after clearing and the absence of clonal spread and resprouting (Jacobs et al. 2014).The incorrect name was perpetuated into Richardson and Rejmánek's (2011) global list of invasive trees and shrubs, but corrected in an update of this list (Rejmánek and Richardson 2013).Although this was not investigated, it is possible that publications citing Melaleuca species from Richardson and Rejmánek (2011) could carry this mistake forward.
Effective pre-emptive control efforts rely heavily on whether alien species are listed as invasive in that region or are known to be invasive elsewhere (Mack 1996).As a result of debate surrounding generic limits in the tribe Melaleuceae, especially regarding the recognition of Callistemon as a segregate genus (Craven 2006;Udovicic and Spencer 2012, Edwards et al. 2010, Craven et al. 2014), species lists included in the recently published Alien and Invasive Species Regulations in South Africa (DEA 2014) may generate errors due to synonymy issues.For example, the regulations list Callistemon rigidus, which is now treated as a synonym of C. linearis if one accepts the separation of the two genera (see Council of Heads of Australasian Herbaria 2016); if a broad concept of Melaleuca is adopted, then the taxon should be listed as Melaleuca linearis.Moreover, several species have been omitted from the regulations, e.g.Melaleuca parvistaminea, a species which is clearly invasive and poses a considerable environmental threat (Jacobs et al. 2014).Recognition of situations like these requires adequate taxonomic expertise and familiarity with the group in question.
Hybridization and horticultural selection for some Melaleuca species, especially those in the Callistemon group can further complicate accurate identification (Brophy et al. 2013).Hybrids and several cultivars exist for some taxa and it is not clear whether some hybrids or cultivars are more invasive than others.Moreover, some Melaleuca species, such as M. linearis, are apomictic and may further contribute to species identification problems.
We identified ten species of Melaleuca naturalised in the Western Cape province of South Africa, but invasions of taxa in this genus are at an early stage, and there is likely to be a high level of invasion debt (sensu Rouget et al. 2016).Unlike other invasive Australian tree and shrub species (e.g.Acacia and Eucalyptus), Melaleuca species were never widely disseminated in South Africa for forestry or dune stabilisation.Melaleuca quinquenervia was introduced and widely disseminated for a variety of reasons, including ecosystem engineering, in the USA (Dray et al. 2006).No wide scale plantings took place in South Africa.Melaleuca introductions and plantings in South Africa have been for ornamental purposes, mostly in the last few decades.Because naturalized populations are still small there is still the opportunity to eradicate several species if action is taken quickly and with sufficient resources.Besides the small populations, other factors that suggest that eradication is feasible are the short-lived serotinous seed banks, the effectiveness of available herbicides (Jacobs et al. 2014, van Wyk andJacobs 2015), limited dispersal capability (inferred from Rejmánek and Richardson 2011) and a focused, national programme with a mandate to respond to incursions (Wilson et al. 2013).The high level of errors in identification which we found in this study, however, highlights the urgent need to assess and improve the accuracy of alien species lists.

Figure 1 .
Figure 1.Examples of the types of errors found on the herbarium specimens examined, a Misidentification: Melaleuca salicina misidentified as Callistemon pallidus b Improved identification: Callistemon sp. was further identified as Melaleuca rugulosa c Synonymy: Callistemon rigidus is a synonym of C. linearis but is currently accepted as Melaleuca linearis var.linearis, and d unresolved taxonomy: prior to 1968, Melaleuca quinquenervia, along with several other broad-leaved species were included under M. leucadendra sensu lato.

Figure 2 .
Figure 2. Examples of naturalized Melaleuca species in South Africa.a naturalized M. quinquenervia plants showing seed capsules opening after fire b M. viminalis subsp.viminalis naturalized along a stream in an urban setting c Melaleuca parvistaminea invading a conservation area that was previously under pine plantation, and d M. linearis var.linearis is invasive at another site previously under plantation with M. parvistaminea in background.Photos: a, c is E van Wyk, b is LEO Jacobs, d is DM Richardson.

Figure 3 .
Figure 3. Localities of naturalized Melaleuca species in South Africa at the resolution of quarter-degree squares (QDS).Darker shading indicates a higher number of species.Grey borders are province boundaries.

Table 2 .
Wilson et al. (2014)11)ecies in South Africa for which there is a confirmed herbarium record in either the Compton Herbarium, Kirstenbosch (NBG) or in the cultivated collection in the National Herbarium (PRE).Note that several other collections were searched, but no additional species could be discovered.Invasive status is according toBlackburn et al. (2011), with interpretation for trees fromWilson et al. (2014).All species were used as ornamentals although older records are often associated with historic forestry sites and arboreta.Later records were sourced mainly from gardens and nurseries.Recently used synonyms are listed and are intended to aid recognition of some species.

Table 3 .
List of Melaleuca species recorded in South Africa for which there is no confirmed herbarium record.