Mapping the extent and spread of multiple plant invasions can help prioritise management in Galapagos National Park

Mapping is an important tool for the management of plant invasions. If landscapes are mapped in an appropriate way, results can help managers decide when and where to prioritize their efforts. We mapped vegetation with the aim of providing key information for managers on the extent, density and rates of spread of multiple invasive species across the landscape. Our case study focused on an area of Galapagos National Park that is faced with the challenge of managing multiple plant invasions. We used satellite imagery to produce a spatially-explicit database of plant species densities in the canopy, finding that 92% of the humid highlands had some degree of invasion and 41% of the canopy was comprised of invasive plants. We also calculated the rate of spread of eight invasive species using known introduction dates, finding that species with the most limited dispersal ability had the slowest spread rates while those able to disperse long distances had a range of spread rates. Our results on spread rate fall at the lower end of the range of published spread rates of invasive plants. This is probably because most studies are based on the entire geographic extent, whereas our estimates took


Introduction
Globally, plant invasions are growing in frequency and areal extent (Mack 2000). These invasions need to be managed because they have serious consequences for biodiversity and the economy (Vilà et al. 2011). Mapping is an important tool for managing plant invasions because it can identify where they are and how long they have been there. Knowing the spatial distribution of invaders can help managers identify sites of invasion (Shaw 2005), monitor the outcomes of management actions (Roura-Pascual et al. 2009) and understand processes that operate at a landscape scale (Richardson 2011). Also, quantitatively documenting the change in areal extent of invasions is important for justifying and sustaining public support of management programs (Mack 2000).
However, not all maps are useful for all purposes. The way the landscape is classified in mapping projects affects the types of management decisions that can be made (Lindenmayer and Hobbs 2007). A traditional vegetation map classifies the landscape into discrete classes that each represent distinct vegetation communities (Küchler 1967). The mapping product may identify the presence of invasive plants as a major or minor element in one or more of the vegetation communities. For example, Garzón-Machado (2011) noted the presence of invasive species in two of the communities in their vegetation map of an island National Park, which could direct the attention of managers of invasive plants to the broad areas of the landscape occupied by those communities. Landscapes can also be classified such that a particular invasive species is a specific focus of the mapping; these can give managers detailed information on the location and dynamics of the species (e.g. Müllerová et al. 2005;Pengra et al. 2007). As plant invasions continue to become more pervasive and all ecosystem researchers/managers are forced to consider invasive species (Richardson 2011), there is an increasing need to map landscapes in a way that takes account of all invasive species that may present.
The methods used to map vegetation, including plant invasions, have evolved over time. The availability of remotely sensed data, especially from satellites, has revolutionised the ability to map vegetation over large areas (Xie et al. 2008). Recently, advanced technology has been used to model invaded and non-invaded forest, giving very specific details on the location, structure and species composition of vegetation containing multiple invasive species (Asner et al. 2008). Whilst this approach provides a plethora of useful information to managers, the application of it is limited due to the expense of obtaining data. In underresourced parts of the world, most managers have access to a limited range of data. As our study was based in the developing country of Ecuador, we aimed to use readily available satellite data to map vegetation in a way that would give managers useful information for managing a landscape containing multiple invasive species.
One of the factors that is important for prioritising the management of plant invasions is the rate of spread of individual invaders (Pyšek and Hulme 2005). Empirical evidence suggests spread is driven primarily by dispersal ability (Coutts et al. 2011). Specifically, the ability to disperse long distances is the main reason for rapid spread of invasive plants (Richardson and Pyšek 2006). Therefore, plants dispersed by wind (e.g. Pinus radiata) or animals (e.g. Acacia cyclops, Opuntia stricta) exhibit the fastest rates of spread, whereas clonal plants tend to have comparatively slower rates of spread (Pyšek and Hulme 2005). Knowing this, managers may choose to delay the management of slow spreaders in favour of managing fast invaders either via attempts to eradicate the species quickly; or adopting a longer term management strategy such as biological control. Information on the rate of spread of invaders can be obtained from the known locations of invasive species at more than one point in time (Gilbert and Liebhold 2010).
Our study focuses on the humid highlands of Galapagos National Park where invasive plants have spread from areas of human habitation (Itow 2003;Rentería and Buddenhagen 2006). Known ecosystem impacts include reduced abundance and diversity of native species (Jäger et al. 2007), which aligns with global concern for the threat of plant invasions on island plant diversity (Caujapé-Castells et al. 2010). The first objective of our study was to map the location, extent and density of invasive canopy species (ecosystem transformers, sensu Gardener et al. 2013;Richardson et al. 2000) in the highlands of Santa Cruz Island, Galapagos. Our second objective was to calculate the rate of spread of individual invasive species. Results from both objectives can help managers decide which species and which parts of the landscape require intervention, and to provide a baseline for monitoring future vegetation change.

Study area
Our study concerns the humid highlands within the Galapagos National Park on Santa Cruz Island, which form a doughnut shape surrounding an agricultural zone and are surrounded by dry lowlands (Figure 1). Non-native plants have been introduced to the island since the first human visitors arrived in the early 1800s, though the majority of the current non-native flora have arrived in the last 30-50 years (Tye 2006). Major plant invasions began in the National Park with the spread of Cinchona pubescens in the 1970s (Eliasson 1982), followed by other invasions from the 1980s onwards (Gardener et al. 2013). To give historical context to the recent plant invasions, we refer to the four most widespread historical vegetation types (HVTs) described and mapped by Trueman et al. (2013): Scalesia Forest, Fern/Herbland & Miconia Shrubland, Mixed Forest and Dry Forest. The extent of these HVTs in the map we produce here differs slightly to that of Trueman et al. (2013) due to the finer scale of data used in this study. Our core area of interest is the humid highlands because they are the most invaded (Guézou et al. 2010) but our study area also includes the transitional periphery between the Figure 1. Location of the study area and density of invasive plants in the canopy. The upper map shows the location of the study area, spanning the humid highlands of Galapagos National Park, Santa Cruz Island. Climatic zones are indicated and the four islands inhabited by people are labelled; each of these is inhabited only in an agricultural zone in the highlands and a small coastal town in the lowlands. The lower map shows the density of invasive plants in the canopy of the study area, as indicated by shading. Co-dominant invasive species may be present with either invasive or native plants. Historical vegetation types are outlined and labelled. Of these the Dry Forest occurs on the periphery of the humid highlands that is transitional to the dry lowlands. humid highlands and the dry lowlands (sensu Trueman and d'Ozouville 2010; Figure  1), coinciding with the HVT Dry Forest.

Mapping the vegetation
We mapped the native and invasive canopy vegetation across the study area using data derived from satellite images and validated with field observations. Canopy vegetation refers to the tallest layer of vegetation, which ranged in height from approximately one metre (e.g., Melinis minutiflora grassland) to over ten metres (e.g., Persea americana forest). Invasive canopy species are ecologically significant because they can modify the structure of vegetation communities and reduce the amount of light penetrating the ecosystem, negatively affecting the abundance of native understorey species and the recruitment of native canopy species (Reinhart et al. 2006).
Mapping involved the creation of a spatially-explicit database. Essentially, we drew polygons over the study region and assigned to each polygon a measure of density of each vegetation cover class using visual assessment of satellite data (Figure 2 and detailed below). We used three separate satellite datasets. Two were Worldview-2 multispectral datasets (2 m resolution, 8-band) as provided by DigitalGlobe. Scene 1 (19 th October 2011, catalog ID: 103001000E276500) covered the western part of the study area and scene 2 (23 rd March 2011, catalog ID: 10300100091E2400) covered the central/ eastern part of the study area. The third dataset was a SPOT 5 pan-sharpened scene (30 th March 2007, 2.5 m resolution, 3-band, level 2A product, image 615/351) which we georectified using 160 ground observation points and a spline transformation in ArcMap 10.0. The SPOT dataset spanned our study area, including small areas in the east that were not covered or were obscured by clouds in the Worldview-2 datasets.
We visually assessed both a true-colour image derived from each satellite dataset and multispectral classifications of each satellite dataset ( Figure 2). Visual analysis of imagery, as traditionally applied to aerial photographs, results in a high degree of map accuracy (Coppin et al. 2004). We used satellite images of sufficiently high resolution (2-2.5 m) to be visually assessed in the way of aerial photographs using features such as colour and texture (Morgan et al. 2010). The added benefit of satellite datasets over aerial photography is that they include reflectance values in the infra-red (non-visible) parts of the light spectrum that are particularly useful for distinguishing different types of vegetation using classification tools (Xie et al. 2008). To take advantage of this, we performed supervised classifications using the maximum likelihood algorithm in ArcMap 10.0 which require human input to select training areas that define a priori classes (Xie et al. 2008). We selected training areas representative of all 26 vegetation cover classes (described below) by visually assessing the true-colour images. We computed several classifications of each dataset because the spectral signatures of the vegetation classes varied over the spatial extent of each dataset. We drew polygons based on the congruence between visual inspection of the true-colour images and the multiple classifications using ArcMap 10.0 at a display scale of 1:5 000 employing the Auto Complete Polygon Tool (Figure 2). Our final database consisted of 1 624 polygons.
In each polygon, we recorded the presence of any of the 26 vegetation cover classes we identified in the images. Twelve of the cover classes represented individual invasive plant species, one class was a mixture of invasive grass species, eight classes represented individual native plant species, three classes represented native plant assemblages, and two classes were non-vegetated (Suppl. material Table 1). The density of each cover classes present in each polygon was scored using the following categories: dominant (60-100% cover), co-dominant (20-50% cover, shared with other species of roughly equal cover summing to a total of 60-100%), secondary (10-20% cover), and scattered (isolated individuals or clusters of individuals with 0-10% cover).
We collected field observations to validate the spatially-explicit database. Data collection points were selected to representatively sample the different patterns visible on the SPOT true-colour image. Field observations were recorded between September-December 2010 and July-November 2011. We recorded the canopy species present within an area of 400 m 2 centred on 591 points within the study area; in total pertaining to approximately 2 000 ha of the full 14 214 ha study area. For validation we intersected the spatially-explicit database with our field observations. The confusion matrix is commonly used for this purpose (Xie et al. 2008) but was not suited to our accuracy assessment because our database contained the density of not one but multiple vegetation classes in each polygon. The spatial intersection confirmed that our database accurately represented field observations as follows: Classes recorded as dominant in our database were observed at 81% of intersecting observation points, and at least one of the two or more species comprising the co-dominant classes was observed at 82% of intersecting points. These scores were consistent with the average estimated total cover represented by these two density categories (60-100%). Classes recorded as secondary in our database were observed at 45% of intersecting points, while classes recorded as scattered in our database were observed in 37% of intersecting points. This is higher than the average estimated cover represented by these density categories (10-20% cover and 0-10% cover respectively), but is consistent with the fact that each of these classes are dispersed within the vegetation cover, and that observation points relate to cover within a small area rather than a single point.

Invasion extent
We mapped the presence of invasive plants in the canopy using the highest density category for any invasive species recorded in each polygon of our spatially-explicit database. For each HVT, we summed the areas of polygons containing invasive plants in these categories. We also calculated the total area invaded by summing the areas of all polygons in which at least one invasive species was recorded. We calculated the approximate total coverage in 2011 of individual invasive species by summing the area of all polygons in which each occurred in each HVT, weighted by the average percentage cover of its density category (i.e., dominant 80%, co-dominant 35%, secondary 15%, scattered 5%). We did this for seven of the invasive species we had mapped as a single class (we excluded Syzygium jambos because it had very low presence) and for the invasive grass species combined. We then calculated the percentage of the canopy vegetation that was comprised of invasive plants by summing the area of total coverage of all invasive species in each HVT and dividing by the total area of each HVT.

Rates of spread
We calculated the mean annual rate of spread of each of invasive species to allow comparison of our results with spread rates reported in the literature (Pyšek and Hulme 2005). Rate of spread is commonly measured in terms of distance per year and can be calculated using a variety of methods (Gilbert and Liebhold 2010;Higgins and Richardson 1999). According to the popular Skellam (1951) model, the square root of the area occupied by an invading organism increases linearly with time (Shigesada et al. 1995). This regression approach is suited to cases such as ours where the introduction location of the invading organism is unknown (Gilbert and Liebhold 2010;Pyšek and Hulme 2005). In our case, the areal extent is only known from a single date, and so we calculated the mean linear rate of spread as the square root of the area occupied at that date divided by the number of years since introduction into the study area (National Park area of Santa Cruz Island). We derived the first record of each species from the literature, herbarium records, and personal communications with past Galapagos plant researchers and residents. Finally, we reviewed literature to compile information on dispersal mechanisms of each canopy invader.

Data resources
The spatial database underpinning the analysis reported are deposited at PANGAEA Data Publisher for Earth & Environmental Science: http://dx.doi.org/10.1594/ PANGAEA.833752

Results
Our spatially-explicit database covered a total area of 14 214 ha, representing the full extent of the highlands of Santa Cruz Island that fall within the Galapagos National Park, including the transitional periphery between the humid highlands and the dry lowlands ( Figure 1). Of this area 7 782 ha (55% of the total area mapped) contained invasive plants in the canopy (Figure 1). Invaders were dominant in 1 527 ha (11%), co-dominant in 1 945 ha (14%), secondary to native species in 1 395 ha (10%), and scattered among native species in 2 916 ha (21%) (Figure 1). Of the invaders present in the canopy as co-dominants, some were co-dominant with native species (especially Cinchona pubescens) while others were co-dominant with both native and invasive species (especially Psidium guajava and Cestrum auriculatum). Overall, 21% of the canopy of the study area was comprised of invasive species.
The Scalesia Forest and Mixed Forest were the most invaded of the HVTs, both with 96% of their area containing invasive plants in the canopy (Figures 1, 3). The HVT Scalesia forest had the highest proportion of the canopy vegetation comprised of invasive plants (52%), followed by the HVT Mixed Forest (46%). Twenty six percent of the canopy of the HVT Fern/Herbland & Miconia Shrubland was comprised of invasive plants. Most of the invasion-free areas occurred in the HVT Dry Forest (Figures 1, 3), of which only 5% of the canopy was comprised of invasive plants. Excluding this drier periphery from our calculations, 92% of the humid highlands contained some degree of invasive plants in the canopy, while approximately 41% of the canopy was comprised of invasive plants.
Of all the invasive plants featured in our database, the first to arrive into the study area was Cinchona pubescens in 1966, and the others arrived later (Table 1). In the 50 year record, Cedrela odorata was the fastest canopy invader and had the largest area of canopy cover (Table 1), predominantly in the HVT Mixed Forest (Suppl. material Table 2). In both speed and extent, this invader was closely followed by Psidium guajava (Table 1 which has spread extensively within all of the HVTs (Suppl. material Table 2). Cestrum auriculatum also had a large area of canopy cover (Table 1), mainly in the HVTs Scalesia Forest and Mixed Forest (Suppl. material Table 2). These three species are the most rapid spreaders of all the invasive species we recorded and are dispersed by wind or birds. However, not all species with these dispersal mechanisms invaded so quickly (Table 1). Persea americana and the grass Pennisetum purpureum had the slowest invasion rates and were the only species purposefully introduced to the National Park and also the only invasive species with vegetative and gravity-assisted means of dispersal (Table 1).

Discussion
Our map of invasions highlights the fact that invasive canopy plants have an extensive distribution in the humid highlands of the Galapagos National Park on Santa Cruz Island. The drier periphery of our study area (the HVT Dry Forest) was less invaded, probably because most invasive plants in Galapagos are suited to wet climates and consequently thrive in the more humid areas (Guézou et al. 2010). We applied a robust, repeatable method of mapping that allows for the comparison of our data with other data that are obtained using similar methods elsewhere or at the same place in the future. Our database details the extent and density of multiple invasive species at a landscape scale and thus provides a benchmark for monitoring future vegetation change. Our map of invasions shows that some areas in the HVT Fern/Herbland and Miconia Shrubland are free of canopy invaders, which is likely a result of management action to control Cinchona pubescens (García and Gardener 2012). Further, our database allows for the identification of areas of high densities of multiple invasive species, where managers might apply control measures for several species, and also areas where careful intervention may be required to conserve the native canopy species that coexist with non-native species. Most of the vegetation classes featured in our database relate to single species, and the density of each is modelled over the landscape. As such, if managers choose to focus on a single species, the percentage cover of that species can be mapped over the landscape. Also, the data can be further classified to match other studies or meet needs for consistency in vegetation classification (De Cáceres and Wiser 2012). The database has already been reclassified to map the distribution of vegetation states of varying degrees of novelty across the study area, to identify options for management (Trueman et al. 2014).
Decisions on management interventions may depend on the rate of spread of individual invasive species. The range of spread rates exhibited by the species in our study can be only partially explained by their dispersal vectors. The two species in our study with the lowest dispersal ability; Persea americana and the grass Pennisetum purpureum -dispersed either by gravity or vegetatively (noting that birds and tortoises also disperse seeds of P. purpureum but seeds tend not to establish (Itow 2003)) were among three of the slowest spreaders (0.003-0.008 km 2 yr -1 ). This result is consistent with research illustrating that spread is driven by dispersal ability (Coutts et al. 2011). However, the remaining species we studied, which are distributed by either wind or animals over long distances, had variable spread rates (0.008-0.35 km 2 yr -1 ). This variability supports the idea that species traits alone do not determine rates of spread (Pyšek and Hulme 2005). Our results fall within the lower end of the range of areal spread rates reported elsewhere, probably because such results are often inflated due to inclusion of the entire geographic extent of species and without density estimates (Pyšek and Hulme 2005). Our study has taken density into account by using the actual area of canopy coverage, and therefore our results are deflated compared with other published rates of spread.
All of the invasions reported in our study have occurred since 1966 or more recently. In less than 50 years 41% of the native canopy vegetation in the humid highlands of Santa Cruz Island in Galapagos National Park has been replaced by invasive species. We have assumed a constant rate of spread, though in reality there is likely to be temporal variability in the spread of invasions (Pyšek and Hulme 2005). In general, plant invasions spread slowly initially (lag-phase), then rapidly (exponential phase), and finally, spread slowly or not at all (Pyšek and Hulme 2005). These phases were apparent for species in our study. For example, Cestrum auriculatum exhibited a lag time of 15-20 years following the first record of its occurrence (1985); it was not recognized as invasive until sometime between (Rentería and Buddenhagen 2006Tye 2001). Conversely, Cinchona pubescens was an earlier invader (first observed in the study area in 1966) that expanded in range quickly and may have reached its peak distribution in the early 2000s (Buddenhagen et al. 2004). It is important for managers to account for potential variability in the timing of invasion among species. For example, nonnative species covering small areas could rapidly expand (i.e., become invasive) after a lag phase. Thus, assuming constant spread rates could lead to management decisions to ignore potential future invasions.
Another important management consideration is that some invasive species requiring management do not feature in the vegetation canopy and are thus not detected in satellite images or data derived from them, such as ours. For example, in our study area Tradescantia fluminensis is a ground-cover plant that has invaded rapidly since its introduction to the study area after 2001 (Fausto Llerena, pers. comm.). By 2011 we observed it widespread in all HVTs except for Fern/Herbland and Miconia, with abundances of up to 100% cover, forming a thick mat that is thought to inhibit the growth of native plants (Gardener et al. 2013). Other species that do feature in the canopy may also require management elsewhere where they only occur in the understorey. Our database featured mono-dominant stands of Rubus niveus that form a canopy, but our methods could not detect where it occurs under other vegetation. During field work in 2011 we observed it as widespread at low to moderate abundance in all HVTs except for Dry Forest, though its extent and density have since grown during years that have been wetter than the long term average (Wilson Cabrera, personal observation 2013). New methods have been used to map the structure of vegetation canopies, producing promising results that will help managers identify invasive species in the understorey and sites of early invasion (Asner et al. 2008).
Future plant invasions are likely in Galapagos. New invasions are predicted to occur from within the existing non-native flora due to the short residence time of many ornamental species and increasing human-mediated propagule pressure (Trueman et al. 2010a). Invasion by current or new non-native species may also be facilitated by the projected increased precipitation in Galapagos (Trueman et al. 2010b). Such a trajectory of ongoing invasions is a huge challenge for management. We reiterate the suggestion of others to apply prevention strategies (e.g. quarantine) and early intervention strategies (e.g. eradicating or containing species that have small distributions) to lower the risk of future invasions (Gardener et al. 2013). Additionally, engagement with private landholders and relevant government agencies is necessary for managing non-native plants in inhabited areas that are the source of invasions to the surrounding National Park.
In summary, invasive plants have been spreading in the last 50 years and now make up a substantial proportion of the canopy vegetation in the humid highlands of Galapagos National Park on Santa Cruz Island. The invasion process is continuing, and early intervention is the strategy most guaranteed to prevent invasion by new arrivals. Local managers can use our spatially-explicit database to identify areas requiring management by targeting sites with multiple invasive species (i.e., site-led management) or particular invasive species (weed-led; Timmins and Owen 2001) and to assess the efficacy of efforts to control canopy invaders, thereby helping to curtail the expansion of current canopy invaders. Our methods could be applied elsewhere to help managers deal with plant invasions across landscapes under their care.

A biography of an invasive terrestrial slug: the spread, distribution and habitat of Deroceras invadens
Introduction Some terrestrial slugs and snails have been inadvertently spread by man well beyond their natural range (Hanna 1966, Barker 1999, Robinson 1999, Cowie 2001, Herbert 2010. They may become very common and cause significant economic damage. For instance, in much of North America the commonest slugs in disturbed habitats are not native species but European slugs such as Deroceras reticulatum (Müller, 1774) and several species of Arion (e.g. Chichester and Getz 1969). These species have also colonised less disturbed habitats. In these cases much of the spread likely occurred well before anyone was observing the process. The current paper gathers data on the spread of the slug Deroceras invadens Reise et al., 2011, which has often been reported (under various different names) as turning up in new places over the last century, thus revealing something of the colonisation process. Many relevant publications are widely dispersed in local journals, so there seems merit in reviewing these accounts in the hope of identifying global patterns. A better knowledge of the colonisation process may help in restricting or slowing the further spread of this and other slug species. The commonest sort of relevant data is the first occurrence in a country or administrative division. Unfortunately this is rather an unreliable statistic, because when a species first arrives its rarity makes its discovery very much a chance event, and because most local malacologists may not recognise the species until its first occurrence in their region has been published. Accordingly, we have tried also to assess rates of spread following the first discovery; unfortunately such monitoring is uncommon.
Our second object is to establish how far D. invadens has spread; several records of our own and unpublished information from museum collections significantly expand the known range. Conversely, some records turn out to be erroneous. Besides the value to those battling pest slugs in affected countries, a fuller knowledge of the range of climates that the species can tolerate may allow us to predict other regions that are at risk of being colonised. Thirdly we are interested in what habitats D. invadens occupies, particularly whether it invades agricultural and natural habitats. A fourth issue is how common the species gets, which can be assessed at various scales, such as proportion of grid squares occupied, proportion of sampling sites at which it is found, or number of animals collected. Here we will often compare with comparable data on D. reticulatum, a usually commoner congener that is found in similar synanthropic habitats but has spread earlier and more widely. Occurrence of D. reticulatum thus provides a proxy for estimating sampling effort (i.e., confirming the activity of someone interested in recording slugs).
The diversity of climates within the introduced range of D. invadens prompts the question of whether several cryptic species might be hidden within the diaspora. Therefore another aim has been to check the species identity of introduced populations. To understand the issues, it is helpful to review the taxonomic background. Deroceras is the largest genus of terrestrial slugs, with over 100 species described (Wiktor 2000). Although the genus is originally Palaearctic, two species, D. reticulatum and D. laeve (Müller, 1774) have spread globally; in the case of D. laeve it is thought that its invasion of America has been natural. Another species, D. sturanyi (Simroth, 1894), is spreading widely in Europe. In much of western Europe and elsewhere in the world malacologists can recognise D. invadens in the field, because it is a different colour to D. reticulatum (although this difference can disappear in alcohol) and larger than D. laeve. However, dissection is required to be sure because D. laeve can sometimes grow large (particularly outside Europe), because D. invadens is externally indistinguishable from D. sturanyi, and because there are other more local species in Europe that are also externally similar, and which also might have spread. In particular, D. invadens was until recently confused with D. panormitanum s.s., a species common in Sicily and Malta (Reise et al. 2011); the vast majority of identifications in the literature predate this splitting. Deroceras caruanae (Pollonera, 1891) is a junior synonym of D. panormitanum (Reise et al. 2011).

Material and methods
Besides checking literature in our own collections, we carried out online searches for "caruanae", "panormitanum" and "invadens" particularly in combination with the names of specific countries. We also searched for the most recent species lists or distribution maps of likely host countries. We checked online museum catalogues, and personally screened the natural history museums in London and Wrocław. A.J. de Winter kindly selected relevant material from the Naturalis Biodiversity Centre, Leiden, and we have also borrowed material from the Rähle collection in the Stuttgart State Museum of Natural History, the Field Museum Chicago, the Florida Museum of Natural History, the Museum Victoria, the Queensland Museum, the Australian Museum, and the University Museum of Zoology Cambridge.
We have incorporated into the account results from our own fieldwork. Specimens collected by HR and JMCH are in the Senckenberg Museum of Natural History at Görlitz (SMNG). DGR has collected separately and specimens are in the collection of the U.S. Department of Agriculture (USDA) at the Academy of Natural Sciences in Philadelphia. Furthermore we have accessed the USDA collection of material intercepted arriving at U.S. ports (Robinson 1999, Reise et al. 2006. M.A. Nash and G.M. Barker provided the SMNG with numerous samples from Australia and New Zealand; other collectors who responded to our requests are listed in the Acknowledgements. HR confirmed identities using characters of the genitalia (Reise et al. 2011). Table  2 lists the collection details of previously unpublished records that extend the distribution of D. invadens significantly. To avoid confusion in subsequent work, we mention identifications of D. invadens in museums or in publications that have proved to be incorrect or were based only on unreliable external characters. Unrecognised misidentifications may affect some of the other records that we analyse or map, but where we had suspicions we have endeavoured to borrow voucher specimens, or we state where further checking would be desirable.
Literature commonly refers to D. invadens as D. caruanae, D. panormitanum or D. pollonerae (older literature also uses the genus name Agriolimax). Moreover, because the separation of D. invadens from D. panormitanum s.s. is recent, much literature is ambiguous to which species it refers. We have found it least confusing here to refer to all such ambiguous records as D. invadens, which is by far the more widespread species, rather than to distinguish unconfirmed records as "D. panormitanum s.l." Table 1 states, for each country or island, how many populations we have confirmed are D. invadens rather than D. panormitanum s.s. Fig. 4A also makes this distinction. The text highlights the very few records outside of Malta and Sicily that were D. panormitanum s.s. This low incidence of D. panormitanum s.s. justifies our working assumption throughout that ambiguous records are of D. invadens. But in no region have we examined enough samples to be confident that some D. panormitanum s.s. are not mixed in; others should continue to check. Table 1. List of countries and oceanic islands dealt with in the text (in the same order). The second column summarises the date D. invadens was first found (outdoors, unless specified). The '≤' symbol indicates when a publication does not give a date of first collection. Dates of interception describe when the species was found on goods derived from that country. The third and fourth columns give the number of sites (or interceptions) for which we are sure that D. invadens rather than D. panormitanum s.s. occurs or vice versa; usually this evidence is our dissections, other cases are from publicly available COI sequences (indicated if this is the only evidence), otherwise the source is cited. We use '-' instead of '0' if there is no evidence of either species.

Results
This section considers each country in turn. Countries are grouped by continent; within continents the ordering is mostly so that geographically close countries are dealt with together; Table 1 can be used as an index to the order of presentation. Oceanic islands are dealt with separately at the end, independently of their political affiliation.

Italy and San Marino
The native range of D. invadens is thought to be in Italy (Reise et al. 2011), but the species might nevertheless be an introduction in parts of that country. Our recent collections and their genetic analysis support this hypothesis, but the results will be published elsewhere. To summarise, we have found that D. invadens is widespread throughout mainland Italy and occurs also in Sicily and Sardinia. It is easiest to find in synanthropic sites, as in other countries, but we also found it in undisturbed woodland. Several other, less frequent, species occur which are externally indistinguishable from D. invadens. Only in one garden in northwest Italy was D. invadens found co-occurring with D. panormitanum s.s. (Table 2); the latter replaces D. invadens in parts of Sicily. Table 2 details the first two records (2013) of D. invadens in the Republic of San Marino (a tiny state surrounded by Italian territory). Reise et al. (2011) argued that Pollonera's (1891) original description of D. caruanae from the capital Valletta referred to D. panormitanum s.s. Our collecting in 2000 on Malta and Gozo (supplemented by that of T. Backeljau in 1994: see Reise et al. 2011) was targeted at Deroceras but never encountered D. invadens; D. panormitanum s.s. and D. golcheri (Altena, 1962) were widespread.

Great Britain
Deroceras invadens was first found about 1930, in Cornwall, but by 1932 also from South Wales, central southern and northeast England, and Scotland (Ellis 1950, Quick 1960, Kerney 1999. This wide distribution implies that it had been overlooked for some time. Formerly, some supposed that D. invadens was native in Britain (Ellis 1951, Hayward 1954), but we are now more aware of how well the species can spread, and further work (Reuse 1983) established that the shell is insufficiently distinctive for us to trust Hayward's (1954) identification of fossils.
The Conchological Society of Great Britain and Ireland publishes lists annually of vice-counties in the British Isles from which species have been newly recorded and the identity confirmed by experts. If we restrict attention to England and Wales (Scotland and Ireland were more sporadically recorded), the number of vice-counties increased only slowly to 13 by 1964, then jumped to 54 within 10 years (presumably at least partly an artefact of increased sampling effort in preparation of a distribution atlas). Over the next 25 years, Kerney (1999) considered that it had continued to spread rapidly. Today, confirmed records are lacking in only one of the 70 vice-counties (for which there are unconfirmed records) and the distribution now spans all of Great Britain, including such offshore islands as Orkney, Shetland, the Western Isles, the Isle of Man, Lundy, and the Scilly Isles.
At a more local scale, D. invadens has been recorded from 1849 10 km grid squares in Great Britain; this is 0.44 as many as the almost ubiquitous D. reticulatum (access of NBN gateway on 24.xi.12; https://data.nbn.org.uk). The distribution map still shows some regions of scarcity (Fig. 1). Some correspond to areas where absent records of D. reticulatum suggest light recording activity, but these are also often areas of low human population density, which itself might have hindered the spread of a synanthrope. Areas in the west with low human population densities nevertheless show a dense distribution of D. invadens. A higher abundance in the west was noted by Kerney (1999) and may derive from the milder climate (both warmer winters and wetter summers). Some regional differences in abundance are certainly not artefacts of recording intensity. For instance, in Suffolk (eastern England, sparsely populated) D. invadens was not reported until 1982 and a particularly thorough survey up to 1990 found it to be relatively scarce, with scattered localities across the county but concentrated in a couple of areas and generally only in gardens and disturbed ground (Killeen 1992). Contrast this with a 1987-88 survey of gardens in Greater Manchester (northwest England, densely populated), which found it to be the slug species occurring in the most gardens (258 out of 372 gardens) and in the greatest numbers (North and Bailey 1989). Similarly, in a diverse sample of 16 gardens in southeast Scotland, D. invadens occurred in 14; D. reticulatum was the only mollusc occurring in more (Sumner 2002). Kerney (1999) nicely summarises where D. invadens is most often found (writing of the British Isles, but it is typical elsewhere too): "A species of disturbed habitats, and associated with roadside rubbish, farmyards and gardens. It often shelters under stones, pieces of wood, cardboard and other litter in bare or sparsely vegetated waste ground". He also reported the species occurring in wilder places such as woods and hedgerows but only in climatically mild areas such as the South West; in our experience this is also the case in the South East. In a survey of eight ancient woods on the Isle of Man, it was found in seven and was the only widespread non-native species (Alexander and Dubbeldam 2013). Likewise, in northeast England D. invadens occurred in 9 out of 17 woods (all contained D. reticulatum : Wardhaugh 1996). Dirzo (1980) reported the absence or rarity of D. invadens in habitats dominated by grass. It occurs in some arable fields (Quick 1949, Foster 1977, Dirzo 1980, Vernavá et al. 2004, Howlett 2005 Chapter 8), where it sometimes dominates, but usually other slugs are the more important pests. However, D. invadens was the dominant slug pest in 10 selected English nurseries growing "hardy nursery stock" (Anon. 2003; the dearth of information about this survey does raise a concern that D. laeve might have been confused sometimes). An ability to thrive in the damp warm environment of modern nurseries would predispose the species to spread into gardens nationwide. Rowson et al. (2014a, b) reported the occurrence of D. panormitanum s.s. from a garden in Cardiff. The single occurrence of this species contrasts with the 37 British populations of D. invadens that we have checked or from which COI sequences appear in Genbank (Table 1).

Island of Ireland
The first records were in 1958, from several sites around Cork, and in 1959 from Newcastle, County Down, at the other end of the island and in a different country (Makings 1959). These dates are surprisingly late, considering the close links with Great Britain and that malacologists familiar with the species there were collecting in Ireland in these 29 years after its discovery in Britain.
Today in Ireland the number of 10 km squares occupied by D. invadens is a similar proportion of those occupied by D. reticulatum as in Great Britain (0.38 vs 0.44; access to NBN gateway on 24.xi.2012; https://data.nbn.org.uk). In an interesting contrast with Britain, Ross (1984) observed that it was commoner to the east and north; this might be explicable by a correlation with higher population densities ( Fig. 1). At least in the north, the explanation for such a correlation is not an absence of collecting activity towards the west (Anderson 1983). Ross (1984) reported that its occurrence in less disturbed sites such as woodlands and marshes was increasing. By 1996, it was "widespread and abundant in disturbed habitats, woodland and marshes throughout Northern Ireland. A notable pest in gardens" (Anderson 1997). Around Cork, even soon after its first discovery, 7-27% of slugs collected in gardens were D. invadens (Makings 1959(Makings , 1962.

France
In 1910 Simroth described Agriolimax scharffi collected in 1903 from La Giandola in the extreme southeast of France. Some (e.g. Bishop 1980, Gavetti et al. 2008) have argued that this refers to the slug we call D. invadens, which would then be the first record of the species in France or anywhere else and give the species name scharffi priority. Unfortunately the description is so casual ("... fand ich neben der gemeinen Ackerschnecke eine kleinere helle Form, die nicht retikuliert, sondern fein dunkel punktiert war": I found together with the common D. reticulatum a smaller pale form which was not reticulated but with fine dark dots) that it could refer to other species known from the area (Bodon et al. 1982), such as D. bisacchium Bodon, Boato and Giusti, 1982, D. rodnae Grossu and Lupu, 1965, or a form of D. reticulatum. Also the sibling species D. panormitanum s.s. occurs only 23 km away at Bordighera, Italy (Table 2). We visited La Giandola in 2013 and did find D. invadens, which in itself is not so informative because the species could well have spread there since Simroth's visit. More significant is that D. invadens at this site is not paler than the D. reticulatum. Conversely some of the D. reticulatum were of a somewhat unfamiliar appearance to us and did fit the description of "pale ... not reticulated ... with fine dark dots". And also in La Giandola itself we found another species fitting this description; anatomically it matched local D. rodnae s.s. but its mating behaviour was distinct. Altogether there seems plenty enough uncertainty to follow Reise et al. (2011) in regarding A. scharffi as a nomen dubium and not to treat this as a record of D. invadens.
The first clear reports of D. invadens from France are from Hameury (1958), listing two sites near Brest (NW France) from 1956. Quick (1960) had examined specimens from the Pyrenées Orientales (SW France). Later Reygrobellet (1963) described a new species D. meridionale, now considered a synonym of D. invadens (Reise et al. 2011); she mentioned it occurring in southeast France (departments of Var, Bouchesdu-Rhône, and Alpes-de-Haute-Provence, without details of localities or dates). More specifically, Chevallier (1973) published an occurrence of D. invadens from this region (Marseille) from 1948. Reygrobellet (1963) also referred to work by Abeloos (1945) demonstrating the existence of two forms of Deroceras near Poitiers (western France), one of which she and Abeloos had subsequently agreed was D. meridionale (i.e. D. invadens). To summarise, the species occurred at two distant sites in France already in the 1940s and was widespread at least by 1963.
A map prepared in 1972 by Chevallier (1973) showed many localities throughout the western half of France and around Paris; their absence on the eastern half of the map, except in the south, was explained by this region not having been surveyed. However Chevallier also stated that the species seemed more at home in coastal regions than inland, and emphasised that it occurred along the very edges of sea shores, estuaries and lagoons as well as by freshwater habitats. In contrast to other habitat descriptions of D. invadens, he mentioned only in the last sentence that it was also sometimes found in gardens.
The first two records from Corsica were from 1977 (Holyoak 1983); these were identified by A. Wiktor but it would be desirable to scrutinise specimens from this island further given the taxonomic complexity of the genus on the adjacent Tuscan archipelago (Giusti 1976). The online database of the Inventaire National du Patrimoine Naturel (INPN) (http://inpn.mnhn.fr/espece/cd_nom/163204/tab/rep) shows D. invadens now to occur also widely in northeast France. Cucherat and Demuynck (2006) describe it as widespread and sometimes locally abundant in the northeast departments of Nord and Pas du Calais. Comparison of INPN maps of D. invadens and D. reticulatum suggests that the continuing absence of records from areas of central and eastern France at least partially reflects an absence of recent recording. However, there certainly are still areas of France where D. invadens is uncommon. Thus, Boulord et al. (2007) found it to be rare to uncommon in Maures (NW France), occurring in only six 5 km squares compared with the 44 with D. reticulatum (ratio = 0.14). And in a review of slugs in Alsace (E France), Hommay (2000) could report D. invadens only from greenhouses near Colmar.

Monaco
We believe that the first record of D. invadens from Monaco is our finding in 2012 under bushes in a park at a spot irrigated by an automatic watering system (Table 2).

Belgium
The first finding of D. invadens in Belgium was in 1968 in a Brussels garden (Van Goethem 1974). The species was not found again until 1972, when a project mapping terrestrial Mollusca began. Van Goethem et al. (1984) mapped on a 10 km grid all findings of D. invadens and D. reticulatum for each year from 1972-83. Again we divide the number of grid squares in which D. invadens had been found by the number for D. reticulatum (thus controlling for collecting coverage); this yields a ratio of 0.65, comparable with figures of 0.44 and 0.38 for Britain and Ireland.
Van Goethem et al. (1984) also displayed the total number of records of D. invadens and D. reticulatum within each grid square. Deroceras reticulatum occurred in almost every square, so for each grid square the ratio of number of records of D. invadens to those of D. reticulatum provides a local measure of commonness of D. invadens adjusted for collecting effort. (Only a small proportion of the records for D. reticulatum predate the occurrence of D. invadens.) By this measure D. invadens is rarer in the southeast (Fig. 2). As in Britain and Ireland, the pattern is a good match at a coarse level to human population density (Fig. 2). A partial exception is the paucity of records around the city of Liege, interesting because in 2012 we failed to find the species in this city despite screening public gardens and allotments that looked ideal habitat.
Additionally the Belgian data provide the number of grid squares in which each species of Deroceras was found each year. Surprisingly, the ratio between the counts for D. invadens and D. reticulatum does not increase over the period 1972-83 (Fig. 3, τ = -0.30, P = 0.19), suggesting that the species was already well established by 1972 despite only one record predating this. Even when we considered only the apparently less saturated southeast, the ratios did not suggest an increase. Our conclusion conflicts with that of Van Goethem et al. (1984), who did sense an increased incidence over this period.
Van Goethem et al. (1984) provided qualitative comments on the habitat of D. invadens. It was more synanthropic than D. reticulatum, and could predominate over that species in such habitats, whereas it was less common on canal banks. It seemed to be absent from the interior of woods, especially coniferous woods, but it was rare even in poplar plantations.  (1984). Size of the square in each 10 km grid square is proportional to the ratio of the number of records of D. invadens to that of D. reticulatum (the very few squares with no record of D. reticulatum show no symbol). This is superimposed on a map of population density (pale green = low, dark blue = high; same source as in Fig. 1).

Netherlands
The first finding was in 1969 in Domburg at the southwest tip of the country (Gittenberger et al. 1970). By 1982 it was recorded from five provinces across the south of the country ( de Winter 1984), and already in 1984 de Winter estimated that he had found it in a third of gardens and similar habitats examined. A distribution map prepared in 2005 (Atlasproject Nederlandse Mollusken; http://www.anemoon.org/anm/ voorlopige-kaarten, accessed 5.vi.13) shows it spread over the whole country, including even the Wadden island of Terschelling (Mienis 2003), although mapping of the Netherlands is much less complete than for the British Isles or Belgium. This map shows D. invadens as recorded in 50 5 km squares between 1990 and 2005, compared with 199 squares for D. reticulatum. The ratio of these numbers is somewhat lower than in Britain, Ireland or Belgium (0.25 vs. 0.44, 0.38 and 0.65), which might partly reflect the lesser sampling intensity in the Netherlands: if each square is sampled rarely the ratio reflects the chance of encountering D. invadens at an individual collecting site, rather than somewhere in the whole grid square. Three intensive local surveys from the south of the Netherlands confirm the broad pattern. In Zuid-Holland, Boesveld (2005a) found D. invadens in 34 5 km squares, compared with 80 for D. reticulatum (ratio = 0.42). For 1 km squares, the corresponding figures are 48 and 129 (ratio = 0.37). In Noord-Brabant, Boesveld (2005b) found D. invadens in 9 of the 118 1 km squares visited, and D. reticulatum in 36 (ratio = 0.25). Both studies pointed out that they had undersampled urban areas, where D. invadens is commoner. Along the coast of Zeeland (Boesveld 2005c), D. invadens occurred in 22 out of the 79 1 km squares visited compared with 38 for D. reticulatum (ratio = 0.58).

Luxembourg
There are four 1997-98 records of D. invadens provided by Weitmann and Groh in the database of the Musée d'histoire naturelle Luxembourg (accessed via http://data.gbif. org/species/5190777 Sept. 2012). Falkner (1979) documents several cases in the 1970s of D. invadens turning up in Munich on purchased lettuce, endive and strawberries, mostly imported from Italy. The first findings outdoors in Germany were in 1977 and 1978 in and near Munich (Falkner 1979). Analysing the subsequent spread is complicated because, although Germany is well supplied with malacologists, recording schemes have been organised, if at all, at the level of the 16 states (Bundesländer). The following paragraphs deal with each state in turn, in order of the first appearance of D. invadens in each, except that the states that were part of East Germany until 1990 are considered in a second paragraph (Berlin is listed with West Germany). The first published occurrence in the former East Germany postdated by a year the opening of the border between East and West Germany (Bössneck 1994), but by this time D. invadens had not been recorded either in 6 of the 11 West German states. Moreover, the notebooks of the late V. Herdam, indicate East German occurrences in Berlin in 1982 and in Brandenburg in 1985 (E. Hackenberg pers. comm.; Table 2); it is hard to judge how reliable these records are.

Germany
Bavaria: first found in 1977 (see above). North Rhine Westphalia: first found in 1979, at several sites near Cologne (Schnell and Schnell 1981); a thorough survey of Cologne between 1990 and 1994 located it at 15 sites throughout the city, compared with 49 for D. reticulatum (ratio = 0.31: Tappert 1996). Baden-Württemberg: first found in 1982, from several distant sites (Schmid 1997). Schleswig-Holstein: first found by 1983 (Wiese 1983); the 1991 atlas (Wiese 1991) showed it in 2 10 km squares, compared with 21 for D. reticulatum (post-1960 records; ratio = 0.10). Hesse: W. Hohorst collected D. invadens in Frankfurt am Main in 1985 (Table 2); in a 2010-11 survey in Frankfurt am Main, D. invadens was recorded from 11 of 22 sites, compared with 5 for D. reticulatum (ratio = 2.2; Kappes et al. 2012). Rhineland Palatinate: first found in 1994, in Jockgrim (Schmid 1997). Lower Saxony: first found in 1998 on the island of Baltrum; other records in Lill (2001) indicate that by 2001 it occurred throughout the state, sometimes at high densities, but restricted to synanthropic sites. Hamburg: first found in greenhouses in 1998 and outdoors in 2000 (Glöer and Haus-dorf 2001). Bremen: first found 2001 (Lill 2001). Berlin: we know of no published record, but note Herdam's 1982 record mentioned above, and we collected it ourselves from a garden in Steglitz in 2001 (Table 2). Saarland: we know of no record.
Saxony: first found in 1990, in Limbach near Reichenbach (Bössneck 1994), and in 1991 at the other end of the state, in Görlitz (Reise and Backeljau 1994). Thuringia: first found in 1993, in Erfurt (Bössneck 1994); rather few sites (c. 5) have turned up since (U. Bössneck,pers. comm. 28.i.14). Mecklenburg-Vorpommern: first found in 1998; by 2006 it had been recorded in 10 5 km squares, always at synanthropic sites (64 squares contained D. reticulatum: ratio = 0.16; Zettler et al. 2006). Saxony-Anhalt: Unruh (2001) stated that it had been part of the outdoor fauna for years but by 2012 it was recorded from only 4 10 km squares (compared with c. 55 with post-1990 records for D. reticulatum: ratio = 0.07; Körnig et al. 2013). Brandenburg: we know of no published record beside our own from 2004, in Senzig at the edge of the Berlin conurbation (Reise et al. 2011), but note Herdam's 1985 record mentioned above.
A 2000-12 survey covering a broad area of northwest Germany, extending also into the Benelux countries, found that the mean density of D. invadens slightly exceeded that of D. reticulatum in gardens and in early successional woodland (Kappes and Schilthuizen 2014) but not in grassland or mature woodlands. The ranking of habitats in decreasing order of mean density of D . invadens was: gardens (0.68 m -2 ), early successional woodland, mesic open, floodplain forest, wet open, wet deciduous forest, deciduous forest, scree forest, conifer; in the later three types the density was very low or absent.
In conclusion, D. invadens is now widely distributed in Germany and can be a common synanthrope in the west, but it took decades to turn up in many states and is still uncommon in many areas. The other invading Deroceras, D. sturanyi, is often commoner (e.g., Zettler et al. 2006, Körnig et al. 2013 although not in the northwest (Kappes and Schilthuizen 2014). There is little indication of D. invadens colonising more natural habitats.

Switzerland
The first record was in Basle near the German border in 1982 (Falkner 1982), the next not until 1991 (Turner et al. 1998). By 2011 D. invadens had still been recorded from only eight 5 km squares (around Basle, Bern, Lucerne and Zürich, so only from the lowlands in the northern half of the country: Boschi 2011). This compares with records of D. reticulatum from 350 squares since 1951 (Boschi 2011; ratio = 0.02). However, an independent study in 2007-09 (Fabian et al. 2012) found D. invadens to be one of the three commonest slugs in wildflower strips edging arable fields south of Lake Neuchâtel; perhaps the mapping had tended to overlook such occurrences in mundane agricultural landscapes.

Austria
The first record was in or before 1977 from a market garden in Maissau (NE Austria; Reischütz 1977Reischütz , 1980. A 1986 atlas (Reischütz 1986) shows D. invadens in 9 10 km squares, all in the eastern quarter of the country, centred on Vienna. However, D. invadens had been collected already in 1979 from Lienz (SSW corner of Austria: Kofler 1986), and in the north also the species has since been found further to the west (1990in Schiebbs: Ressl 2005 in Braunau am Inn: collected F. Seidl, Table 2; 2009 in Linz, 2011 in Wels: Aescht and Bisenberger 2011). In 2013 we found the species still further west, in Innsbruck (Table 2).
A 1992 survey of greenhouses around Vienna found D. invadens in 3 out of 10 establishments (the same as for D. reticulatum; D. laeve occurred in 6: Leiss and Reischutz 1996).

Czech Republic
A specimen collected in Ostrava in 1996 was identified as D. invadens only in 2003 . By that time there was a further 2002 record from a greenhouse in Sušice. Horsák and Dvořák (2003) believed that it was liable to have been widely overlooked elsewhere, but the latest review (Horsák et al. 2013a) provides no further records. We have recently (2014) found the species at two sites within the town of Hrádek nad Nisou, close to the German and Polish borders (Table 2).

Slovakia
Deroceras invadens is known only from greenhouses of the botanical garden in Bratislava where it was first found in 2003 Horsák et al. 2013a).

Poland
The only record of which we are aware is from 2001 from the botanical garden and adjacent areas in Wrocław (SW Poland: Wiktor 2001aWiktor , 2004. The population was still extant in 2013 (A. Wiktor and HR pers. obs.).

Lithuania
Deroceras invadens had been reported from the botanic garden in Kaunas (Skujienė G 2013 Invasive slugs in Lithuania: results, problems and perspectives of the investigations. Abstract booklet of "Slugs and snails as invasive species, a meeting of the IOBC/ WPRS slugs and snails subgroup, Bergen, Norway 25-27 September 2013", p. 11). However, Skujienė kindly lent us the specimen on which this record was based and it proved to have been misidentified.

Hungary
Deroceras invadens was listed as present in Hungary in the guide book by Kerney et al. (1983). However, the more authoritative lists of Wiktor and Szigethy (1983) and Pintér and Suara (2004) do not include this species. Grossu (1969) retracted the earlier claim that D. invadens occurred in Romania (Grossu and Lupu 1965).

Bulgaria
Wiktor (2000) mentioned a population in the Piryn mountains of Bulgaria that he assigned to D. panormitanum. At that time the confounding of D. panormitanum and D. invadens meant that the genitalia of the species were considered very variable. Now we reject this identification; for example, the sarcobelum is strongly flattened in these Bulgarian slugs, which is utterly untypical of D. invadens, or of D. panormitanum s.s. Wiktor (2001b) reported "Deroceras cf. panormitanum" at "scattered localities" in Greece; he considered the 202 specimens from many sites to be all one species but was tentative in placing them in this taxon. However, specimens fitting this description from Crete proved to be an undescribed species (Reise et al. 2011). We have now checked the anatomy of five Greek samples in the Museum of Natural History, Wrocław University (from Mount Lampeia in the Erymanthos mountains, Cape Sounion south of Athens, the island of Skyros, the Kallicratis plateau in SW Crete, Heraklion in Crete) and one sample in the Naturalis Biodiversity Centre, Leiden (RMNH.MOL 329848 from the island of Lefkada), all labelled by Wiktor as D. panormitanum. In the light of the recent taxonomic revision (Reise et al. 2011), our opinion is that none of these are D. invadens or D. panormitanum s.s., so we concluded that probably all Greek records of D. invadens should be considered unreliable.

Greece
However, recently Rowson et al. (2014a) have published a finding of D. invadens from Lake Kournas in Crete reliably identified by a COI gene sequence (Genbank KF894343; R. Anderson pers. comm. 15.iv.2014). Anderson had found the species in a patch of disturbed scrubland in May 2011 and 2012. Crete has been fairly well sampled for slugs without this species being noted elsewhere, so a recent introduction seems likeliest.

Denmark
The first specimens were already collected in 1937, outdoors in a park in Odense. But their identity was not recognised until the mid 1950s when further specimens turned up at seven different cemeteries in another part of Denmark (northern end of Jutland: Lohmander 1959). We have found no more recent information.
See below for records from the Faroe Islands.

Sweden
The first findings date from 1957 to 1959, when D. invadens was found in six greenhouses well spread over the country (even up in Bysek at 65°N; Waldén 1960). However, subsequently the species remained rare in greenhouses or even declined (Proschwitz 1991). Outside greenhouses, occurrences also remained "extremely rare" before 1980, but by 2002 D. invadens was rapidly establishing itself, perhaps as a result of milder winters (Proschwitz 2002). It is now recorded outdoors from the provinces of Skåne, Halland, Småland, Västergötland, Dalsland, Uppland, Värmland and the Baltic islands of Öland and Gotland, with records denser in the south but extending as far north as Uppsala (59. 8°N: Proschwitz 20098°N: Proschwitz , 2010.

Norway
The first find was in about 1967 from a greenhouse in a botanic garden in Bergen (Olsen 2002). Outdoors, D. invadens was first found in 1983-84 at four synanthropic sites along the mild Atlantic coast, as far north as Møre og Romsdal (c. 62°N: notebooks of H.W. Waldén, cited in http://databank.artsdatabanken.no/FremmedArt2012/N79713 accessed July 2013). By 2002 it was considered common outdoors in cultural habitats (Olsen 2002), and it is spreading to more natural habitats (Sneli et al. 2006). Brander and Kantee (1961) reported D. invadens from Finland, without stating the date of discovery. Valovirta (1967) made clear that it was known only from the greenhouses of southern Finland. We are not aware of more recent information.

Finland
Portugal Seixas (1978) was the first to report D. invadens, from near Lisbon in 1977. Rodríguez et al. (1993) reported three further localities in the northern half of Portugal from 1983-85 (out of 55 sites visited across Portugal, 40 of which yielded D. reticulatum: ratio = 0.05). See below for records from Madeira and the Azores. The indications of D. panormitanum s.s., together with D. invadens, on both these archipelagos suggest that it would be worthwhile rechecking specimens from mainland Portugal.

Spain
The first published record of D. invadens was from Bilbao (north coast) in 1980 (Gómez et al. 1981). However, Castillejo (1983) had identified material collected as early as 1974 from northeast Spain; over the period 1974-80 D. invadens had been found in 16 10 km squares in Galicia (compared with 36 for D. reticulatum: ratio = 0.44; records added over the next decade changed this to 20:50 = 0.40: Castillejo and Rodríguez 1991). Later mapping (Castillejo 1997) indicates a distribution along the wet northern coast, with one record in the southern foothills of the Pyrenees ENE of Huesca and some around Valencia (middle of east coast). The species is rarer in the east than in the northwest: in the east-coast province of Castellon, Borredà and Collado (1996) found D. invadens in 2 out of 105 locations sampled in 1990-92, compared with 61 for D. reticulatum (2 vs 38 10 km squares occupied by each species: ratio = 0.05). Similarly in the neighbouring east-coast province of Valencia, D. invadens occurred in 2 out of 69 localities sampled in 1990 and D. reticulatum in 34 (2 vs 27 10 km squares: ratio = 0.07; Borredà et al. 1990). Two of our own findings (Table 2) extend the range to the northeast (Bàscara = first record for Catalonia) and further south of the Pyrenees (Tudela). There remains a dearth of records from the drier middle and south of the country. Survey work in Castilla-La Mancha (central Spain) 2003-07, located only one site for D. invadens compared with about 101 for D. reticulatum (Bragado et al. 2010).
The first record for the Balearic Islands was 2001 from Majorca, where D. invadens has since been found in several other localities (Beckmann 2007). There are also records from Ibiza in 2001 (under pots in garden centre: Anderson 2003) and from Menorca in 2002 (hedgerows and/or woods: Anderson 2004). Castillejo (1998) emphasised that D. invadens is restricted to synanthropic habitats in mainland Spain, especially gardens, whereas Altonaga et al. (1994) listed wet meadows and beside streams. In northwest Spain it occurs in arable fields and permanent cattle-grazed pasture, in both of which it is sometimes the commonest slug (Iglesias et al. 2001(Iglesias et al. , 2003Cordoba et al. 2011).
See below for records from the Canary Islands. . This was the first report from Egypt, probably partly because the authors sent some specimens abroad to be identified.

South Africa
Deroceras invadens was recorded from several widely separated sites in Cape Province (George, Wilderness, Cape Town) in 1963-65 (Altena 1966). Now (Herbert 2010), it is also known from the Eastern Cape and from Guateng, including Pretoria, which, at 25.7°S, has a humid subtropical climate (cf. the mediterranean climate of the Cape).
Although most records are from gardens it has been found also in indigenous forest on

Kenya
A specimen was intercepted arriving in the USA on cut flowers (Astrantia) from Kenya (04.viii.12, USDA 110834). One area in Kenya where Astrantia is grown commercially for export is at Kipipiri, at an altitude of 2300-3000 m. Such altitudes may well provide a suitable habitat for D. invadens.

Sri Lanka
Two publications report the presence of D. invadens in Sri Lanka (Kumburegama andRanawana 2001, Bambaradeniya 2002) and this information has been repeated in secondary sources. However, a later review paper including some of the same authors, but in conjunction with European experts, did not list the species (Naggs et al. 2003). The explanation is that the original reports were erroneous (N.P.S. Kumburegama,pers. comm. 21.viii.13).

Australia
Altena and Smith (1975) were the first to identify D. invadens in Australia. The earliest collection date of the specimens they identified was 1967, although in 1964 specimens had been intercepted arriving in New Zealand on plants from Australia (Barker 1979 1975, Smith 1992a, Stanisic et al. 2010, Atlas of Living Australia http://bie.ala. org.au/species/Deroceras+panormitanum accessed 21.vii.13). However, we have borrowed some of the material in the Australian museums on which these records were based and found that much of it is misidentified D. laeve. As in North America, D. laeve in Australia often grows larger than in Europe, which might have misled originally. Altena and Smith (1975) did not encounter D. laeve, but did not dissect material from outside Victoria that they assumed to be D. invadens. The belief that D. laeve was rare in Australia and, for instance, did not occur at all in Queensland (Stanisic et al. 2010) must have inhibited routine dissection. The latest Australian guidebook gives general coloration and paleness around the pneumostome as characters to distinguish D. laeve and D. invadens (Stanisic et al. 2010); these are certainly not reliable characters in Europe and may well explain misidentifications in Australia.
We can confirm the occurrence of D. invadens in Victoria, New South Wales, Tasmania, South Australia and Western Australia (Table 2). There is no reliable evidence of D. invadens in Queensland. We have examined 14 lots covering a wide geographic range in Queensland from the New South Wales border as far north as 16° (10 labelled as D. panormitanum from the Queensland Museum, one similarly labelled from the Museum Victoria, and 3 others). All were D. laeve. The northernmost confirmed records of D. invadens in New South Wales are 30.4°S (Table 2).
Already in the 1975, D. invadens was described as "one of the commonest and most wide-spread of introduced slugs", "a pest both of pasture plants and those of suburban gardens", occurring "in only slightly disturbed native bushland as well as wholly modified habitats" (Altena and Smith 1975). More recent surveys found it to be the commonest introduced mollusc in native grassland in southeast Australia (Holland et al. 2007), and to occur also around the edges of ponds and swamps (Stanisic et al. 2010). It is found in some arable fields (Nash et al. 2007). The frequent misidentifications of D. laeve as D. invadens might have coloured these claims, but we can confirm its presence in gardens, agricultural habitats, and rough grassland.
See below for erroneous records from Lord Howe Island and Norfolk Island. Barker (1979) published the first records of D. invadens from New Zealand, including specimens collected in 1974. Barker (1979Barker ( , 1999 concluded that earlier widespread records of D. laeve must have referred to D. invadens, because he found that D. laeve had a restricted distribution in New Zealand. For instance, Suter (1913Suter ( , p. 1071 described D. laeve as cosmopolitan, and already in 1891 Musson mentioned that D. laeve occurred. An occurrence at either of these dates would predate the first records of D. invadens anywhere in the world. The argument mirrors that for an early occurrence in Australia, which we criticised above, but we find it more compelling in the case of New Zealand: New Zealand has been more thoroughly surveyed, the distribution of D. laeve was more restricted at the time of these reliable surveys, and, unlike in Australia, there are no extant specimens of D. laeve collected before the late 1950s. Barker (1999) described the habitat of D. invadens in New Zealand thus; "a slug of moist habitats found in gardens, parks, and pastures, on arable land, stream banks, and roadsides, and in greatly disturbed areas of native forest". It is "frequently found in roadside margins adjacent to forest, often kilometres from any cultivated areas", "ubiquitous where there is improved grassland and cultivated plants", "often the most abundant species in humid glasshouses and wet field situations", and "an important pest of cultivated plants" (Barker 1982(Barker , 1992. In one old pasture, 39% of slugs were D. invadens (= 14 m -2 , cf. 16 m -2 D. reticulatum; Barker 1990). Although most often recorded in the Auckland area of North Island, D. invadens also occurs widely around the coast of South Island and on Stewart Island (Barker 1999).

New Zealand
See below for records from Raoul Island and the Chatham Islands.

Americas
United States of America Pilsbry (1948) described D. invadens as widespread in the Bay Area of California (the environs of San Francisco) by 1940; habitats he listed were a park, nurseries, a lawn and a ranch. Subsequently it has been recognised further south, in Monterey County (Lange 1944, Pilsbry 1948) and in Los Angeles (Roth and Sadeghian 2003). It also occurs further north, in Oregon and Washington: Burke (2013) described it as common in urban and suburban areas of the Pacific Northwest from California to southern British Columbia and Pearce et al. (2013) described it as well established in Olympia and Tacoma (Washington State). Table 2 includes further records from this region.
With this firm foothold , one would expect the species to have become widely distributed in the USA, because some large horticultural firms grow plants in the benign climate of the Pacific Northwest and then ship nationally. However, compared with Europe, North America has far fewer experts able to identify slugs, especially those requiring dissection. Also problematic is that D. laeve in North America often grows larger than in Europe so that D. invadens is readily mistaken for it unless specimens are dissected (Reise et al. 2006). It is not a healthy sign that all the records of D. invadens from the USA outside of the west coast are our own! In 1998 we found D. invadens in Washington DC, under bushes in a park, the first outdoor record in eastern North America (Reise et al. 2006). Our surveys of synanthropic habitats in Colorado and Utah in 2004 and 2006 respectively (JMCH and HR, unpublished; see Table 2) turned up D. invadens in several garden centres and in the watered plantings in city centres (e.g. Park City; Memory Grove Park, Salt Lake City: Reise et al. 2011), but also in unwatered parts of a city park (Cheesman Park, Denver), along the banks of a drainage ditch (university campus, Fort Collins) and beside a stream in a ski resort (Snowbird).
The only other records from the eastern USA of which we are aware are from two sites in Kentucky (specimens in Florida Museum of Natural History, catalogue numbers 43778 and 44718, details available via http://data.gbif.org/species/5190777 accessed 24.iii.14). However, we have dissected one animal from each sample and they were not D. invadens.

Canada
The first Canadian records are from greenhouses in two cities in Quebec Province in 1966 (Chichester and Getz 1969). This thorough survey of 770 sites (although only 25 greenhouses, nurseries or gardens) in the northeast of the USA and southeast of Canada failed to find the species elsewhere, whereas D. reticulatum and other European slugs were frequently encountered. The first outdoor records are from the other side of the continent on the UBC campus in Vancouver in 1974 (Rollo and Wellington 1975). Deroceras invadens is nowadays present at other sites in Greater Vancouver and around Victoria on Vancouver Island (Forsyth et al. 2001, Forsyth 2004, http://linnet. geog.ubc.ca/efauna/Atlas/Atlas.aspx?sciname=Deroceras%20invadens; Table 2). We also found it in 2013 in a garden centre in Kamloops, in the drier interior of British Columbia (Table 2). On the Atlantic coast, D. invadens was found in 2012 on rough ground adjacent to gardens in St John's, Newfoundland (Forsyth 2014; confirmed by HR); the maritime influence there ameliorates winter temperatures.
Identifications of D. invadens from a garden in Edmonton, Alberta and from orchards by Osoyoos Lake, British Columbia, (Neckheim 2013(Neckheim , 2014 were not based on dissection (C.M. Neckheim pers. comm. 02.x.13), so should be regarded as unconfirmed; winters in Edmonton are more extreme than in any area where the species is known with certainty (see Discussion). A 1994 record from wild habitat near Kingston, Ontario (Grimm et al. 2009) is also best considered as unconfirmed because of other misidentifications by the collector (Forsyth 2013).

Mexico
We have confirmed the identification of specimens of D. invadens collected by A.S.H. Breure in 1974 at 3000 m in the Desierto de los Leones National Park, above Mexico City ( Table 2). The gardens of the nearby convent might have been the original point of introduction but these slugs were collected in the surrounding pine and oak woodland (A.S.H. Breure 1974, pers. comm. 27.ix.13). At this altitude the climate is cool and damp. Although the specimens were correctly identified by Altena and deposited in the Naturalis Biodiversity Centre, Leiden, the record was not published so that D. invadens is missing from a recent national checklist (Thompson 2008).

Costa Rica
In 2006, we found ten specimens under rocks in a small wood near Tierra Blanca, Provincia Cartago (USDA 131032; Table 2). The altitude was 2060 m. The only Deroceras species listed from Costa Rica by Barrientos (2003) was D. laeve.

Panamá
In July 2007, a specimen of D. invadens was found on a leaf imported into the USA from Panamá (USDA 131034). In July 2009, three further specimens were found on Allium imported into the USA from Panamá (USDA 131033). Note that Panamá connects Costa Rica to Colombia, countries for which the presence of D. invadens has been confirmed on the ground.

Colombia
Two specimens of D. invadens in the Field Museum Chicago (JK-198690, identity confirmed by HR) were collected by the University of Oxford expedition to Colombia in September 1975. There are no further locality data but in this month the expedition was both near Nazaret (Guajira state) and in the capital Bogotá (Knappett et al. 1976).
Deroceras invadens was next found in 2000 at two rural sites near Bogotá in a garden and a flower plantation (Hausdorf 2002). It was considered a serious horticultural pest. Although the localities are less than 5°N, the altitude (> 2600 m) makes the climate oceanic (subtropical highland). Until Hausdorf's visit, the species had apparently not been recognised by local researchers, so that it may be more widespread. However, D. invadens was not found at 6 of the 8 sites where Hausdorf's survey found D. reticulatum (ratio = 0.25).
Cut flowers imported into the USA from Colombia in March 2008 contained D. invadens (USDA 131036).

Ecuador
In April 2012, L. Manangón collected a specimen "on alder" near Bolivar, Provincia Carchi (USDA 110614: Table 2). Although this is only 56 km north of the equator, the altitude of 2600 m meliorates the climate (cf. the localities in Colombia, 600 km away along the chain of the Andes). Earlier, in August 2004, an individual of D. invadens had been intercepted on cut flowers imported into the USA from Ecuador (USDA 131035).

Peru
The only Deroceras species listed from Peru by Ramírez et al. (2003) were D. laeve and D. reticulatum. However, in March 2012, an individual of D. invadens was intercepted on lettuce imported into the USA from Peru (USDA 110831). Peru connects Ecuador and Chile, countries for which the presence of D. invadens has been confirmed on the ground.
Chile Letelier et al. (2003) reported D. invadens as present in southern Chile. We can add three later records of our own from central Chile in, and within 70 km of, the capital Santiago (Table 2). At these latitudes, the species occurs at low altitudes.
See below for a record from the Juan Fernández Islands. The presence of D. invadens there in 1962 suggests that it was probably present in mainland Chile by this time. Brazil Barker (1999) listed two samples of D. invadens that he examined, from a park in the city of Porto Alegre and from a state park 70 km outside the city, both collected in 1991. These determinations should be considered reliable because Barker was familiar with D. invadens in New Zealand. However, recent species lists from the same state (Agudo-Padrón 2009) and from all Brazil (Simone 2006, Agudo-Padrón and Lenhard 2010) do not mention this species. They do list D. laeve, which seems likely to have masked the presence of D. invadens from local malacologists.

Argentina
Recently (27.ii.14) a specimen of D. invadens was intercepted arriving in the USA on cabbage from a ship's stores that had been loaded in Brazil (USDA 140148).

Faroe Islands (Denmark)
This sizeable archipelago (1400 km 2 , population 50,000) lies between Scotland and Iceland at 62°N, having a maritime subarctic climate. In 1970, McMillan (1972 found D. invadens to be frequent in waste ground to the west of the capital Tórshavn.

Madeira (Portugal)
The first record of D. invadens is from 1980 (Rähle 1992). Based on fieldwork from 1980 to 1986, Rähle (1992) concluded that it was widespread but rare and restricted to synanthropic sites on the main island. Seddon (2008) provided more records, but the species had still not been recorded from Porto Santo or other minor islands. Whereas Rähle knew it from 7 sites compared with 10 for D. reticulatum (ratio = 0.7), Seddon's distribution maps show 21 dots for D. invadens and 11 for D. reticulatum. However, Seddon's data may have confounded occurrences of D. invadens with those of Deroceras lombricoides (Morelet, 1845): Seddon has no records of the latter species, whereas Rähle found it to be the commonest Deroceras. Rähle's (1992) illustrations indicate that he found D. invadens rather than D. panormitanum s.s. We have also confirmed the identity of samples in the Stuttgart State Museum of Natural History from 6 localities (collected 1980-85) and of samples in the Naturalis Biodiversity Centre, Leiden, from a further 11 localities (collected 1987-88). However, we collected D. panormitanum s.s. from the south of Madeira in 2006 and again eight years later just 0.7 km away (USDA 131030, USDA 140106: Table 2); the localities lie <2.5 km from one where D. invadens was found in 1988. Waldén (1960) mentions the first finding of D. invadens in the Azores by a 1957 expedition (see Backhuys 1975, p. 22). However, Backhuys' (1975) team of collectors did not find this species during their extensive fieldwork in 1969, in contrast to numerous records for other slugs, including D. laeve (reported as always aphallic, so not a misidentification). Backhuys (1975) reported a single later finding in 1974 from a synanthropic habitat on São Miguel. In dramatic contrast, single-island surveys 13-20 years later showed the species to have become widespread (e.g., present at 5 out of 29 sampling stations on São Miguel in 1987, 8/21 stations on Flores in 1989, 5/17 stations on Santa Maria in 1990, 2/9 stations on São Jorge in 1992, 3/18 stations on Faial in 1993: de Winter 1988;Martins et al. 1990Martins et al. , 1991Martins et al. , 1993Cunha et al. 1994). By 2010 the species was known from all of the nine main islands except Terceira (Cunha, Rodrigues and Martin 2010).

Azores (Portugal)
A slug intercepted arriving in the USA from the Azores on a taro root in March 2008 was D. panormitanum s.s. (USDA 110434). So we checked specimens collected by J. Wieringa from four sites on São Miguel in 1987 (Naturalis Biodiversity Centre, Leiden: collection numbers 329842-329845). They were D. invadens, as was another U.S. interception from the Azores (USDA 131029, Dec. 2008). Probably, as on Madeira, both species occur.

Canary Islands (Spain)
The first records of D. invadens are from the island of La Palma in 1947 (Altena 1950). In the 1980s Alonso et al. (1986) found it to be commoner there than D. reticulatum (12 vs 2 localities, 8 vs 1 5 km squares) and to range in altitude from 60 to 1800 m. On the island of Tenerife, Altena (1950) failed to find D. invadens in 1947 and extensive sampling in 1982-85 located it at only one site (Alonso et al. 1986), but by 2007 it was widespread in native laurel forests (Kappes et al. 2009). The species was also present on Gran Canaria by 1984 (M. Ibáñez, pers. comm. 16.ix.13, Table 2).

Tristan da Cunha (UK)
This lies in the middle of the South Atlantic (37.1°S), 2816 km from South Africa, with a population of under 300. The climate is temperate. Preece's (2001) survey in 1982-83 found D. invadens not only on the main island of Tristan da Cunha but also on Inaccessible Island 40 km away. The latter had been inhabited rarely, and not since 1938 (Anon. 2011). Seabirds abound, so possibly they could have dispersed slugs from Tristan da Cunha.

Raoul Island (New Zealand)
This lies 29.3°S in the South Pacific, 1100 km NNE of New Zealand's North Island. The climate is subtropical. Deroceras invadens was found in 1973 in forest litter (Barker 1999). At this time there was a small farm on the island as well as a meteorological station, although resupply was infrequent.

Chatham Islands (New Zealand)
These lie 44°S in the South Pacific, 680 km from New Zealand. The climate is temperate and they have a sizeable agricultural community and frequent transport links. Deroceras invadens was found in 1976 in pasture on the main island and nearby Pitt Island (Barker 1979).

Marion Island (South Africa)
This is a subantarctic island (46.9°S, 290 km 2 ) with a cool oceanic climate, unpopulated except for research stations; South Africa lies 1730 km to the northwest. Deroceras invadens was first reported in 1972, under timber and in damp mossy habitat beside the base hut; a thorough survey in 1965-66 had not reported it (Smith 1992b). By 1976-77 it occurred in further habitats but had not spread far, and by the early 1990s it was much more abundant but still known only within several hundred metres of the base (Smith 1992b, Chown et al. 2002. Further dispersal around the rest of the island over inhospitable terrain is thought to have been facilitated by its habit of sitting under wooden boxes lying on the ground, which helicopters then carried to other huts (Chown et al. 2002). The species has now spread right around the island but the cold restricts it to land under about 200 m (Lee et al. 2009). It is most abundant near the coast.

Juan Fernández Islands (Chile)
In the Field Museum Chicago and the Museum of Natural History, Wrocław University are specimens of D. invadens collected in 1962 from Robinsón Crusoe Island, formerly Más a Tierra (identities confirmed by HR; Field Museum catalogue number = 198633). This Pacific island is 48 km 2 in area, and lies 600 km west of mainland Chile, with a mediterranean climate and a population in 1999 of over 500. The slugs were collected in a ravine in the Valle de Lord Anson, which rises from the main village.

Errors and Absences
Lord Howe Island and Norfolk Island are small (15 and 35 km 2 ) but well populated Pacific islands, belonging to Australia although 570 and 1400 km east from the Australian mainland. Online records from the Australian Museum (http://ozcan.ala.org. au accessed 08.iii.2013) indicated that D. invadens was collected on Lord Howe Island in 2000 (single record) and on Norfolk Island in 1999 and 2002 (6 records, most from the largest patch of woodland but two from the opposite side of the island). We have borrowed the specimen from Lord Howe Island and two from Norfolk Island (one from each year); one from each island were D. laeve and the third specimen was not identifiable. Given the high rate of similar misidentifications of other Australian material (including from this museum; see above), we consider that there is no reliable evidence of D. invadens occurring on either island.
Specimens in the Natural History Museum London labelled as D. caruanae from São Tomé in the Gulf of Guinea (collected 1993 by A. Gascoigne from Lagoa Amelia and Tras-os-Montes; BMNH 19991797, 19991798) turned out to be D. laeve (dissection by HR); these records appear not to have been published or put online.
It may also be helpful to list some oceanic islands where D. invadens has not been found even though the climate might be suitable and recent surveys have been extensive and informed enough to have probably revealed the species were it well established: Iceland (Sumner 2007), Bermuda (Bieler and Slapcinsky 2000), Cape Verde Islands  invadens on a grid of one degree of latitude and longitude; exceptions are small oceanic islands (single symbol for each island group) and when records specify only a region which overlaps the grid lines (California, N Norway, Öland, and interceptions; a single symbol is marked in a representative "square"). Green cross = records only from greenhouse or garden centre; orange or magenta square = any other record (including garden or park); magenta indicates that at least one record has been confirmed to be D. invadens rather than D. panormitanum s.s.; circled i = only evidence of presence is interception on produce exported from that country. Swedish records in Waldén (2007) are taken to be from greenhouses. In regions where there has been taxonomic confusion (Italy, Balkans, Australia), we have excluded records not verified by ourselves (exceptions are a COI sequence from Crete, and data in Barker (1992) and Altena & Smith (1975) confirmed by dissection) B The squares in A are replaced by black dots, and these are superimposed on a map of the most relevant Köppen climate categories, as modified and interpolated by Peel et al. (2007). Cb + Cc = temperate with warm or cold summer (T h < 22 °C, 0 °C < T c < 18 °C). Ca = temperate with hot summer (T h > 22 °C, 0 °C < T c < 18 °C). Da + Db = cold winter, with hot or warm summer (T c < 0 °C, T h > 22 °C or N T>10 ≥ 4). BSk = cold steppe (5 P T < P < 10 P T , T m < 18 °C). T h = temperature of hottest month, T c = temperature of coldest month, T m = mean annual temperature, N T>10 = number of months when temperature is above 10 °C, P = annual precipitation, P T = constant set by T m and timing of precipitation C The squares in A are replaced by black dots, and these are superimposed on a map indicating winter temperature. For each calendar month, the mean of the daily minimum temperature was calculated, this was averaged over years, and the lowest value amongst the calendar months used (Hijmans et al. 2005; http://www.worldclim.org/ accessed 01.x.2013). Maps created in QGIS 2.0.1 (QGIS Development Team 2013) using outlines from Natural Earth.

Habitat
In Europe much of the area occupied by D. invadens counts as temperate according to the Köppen climate scale (Fig. 4B). This holds also for much of its range outside of Europe. Further areas of the world having this climate category, such as the southern USA and much of China and Japan, would be worth checking further for the presence of D. invadens, and should consider themselves at risk of invasion. Moreover, the presence of D. invadens in such regions as Austria and Newfoundland demonstrates that it can exist in areas with somewhat colder winters than allowed under Köppen's temperate categories (C), whereas its presence in eastern Spain shows that it can extend into drier climates classed as steppe (BS). Conversely, even around the Mediterranean, D. invadens does not occupy all the region classed as temperate. Inevitably, the criteria of the Köppen scale do not exactly match the critical factors defining the species' niche. Cold seems to be one critical factor. In the laboratory, D. invadens collected from Marion Island was unable to survive brief temperatures lower than -6.4 °C, or, on a longer time scale, temperatures lower than about -3 °C on average. This neatly explained its altitudinal range on that island (Lee et al. 2009). Moreover, the gradual decline in density with altitude is explicable by a gradual reduction in performance with decreasing temperatures (Lee et al. 2009). Analogously, the slowness to penetrate Eastern Europe perhaps has as much to do with the cold winters as the restrictions on east-west trade until 1990. Temperatures experienced by slugs in winter depend on snow cover and how deep the slugs burrow underground, but weather stations normally measure only air temperature. In Europe, the coldest locations for which D. invadens has been recorded outdoors are Ultuna near Uppsala in Sweden (winter minimum air temperature averaged over last 20 years = -22 °C; coldest over same period = -29 °C), Södra Fjöle in Sweden (-22 °C and -28 °C for these statistics), and Ostrava in the Czech Republic (-19 °C and -27 °C: Klein Tank et al. 2002, http:// www.ecad.eu accessed 25.ix.13). For Wrocław, we know the species has survived the last 12 winters, in three of which air temperatures dropped to -22 °C. Figure 4C suggests that winter temperatures could well block further expansion eastwards from Wrocław or northwards from Uppsala, but there does seem an opportunity for further expansion elsewhere, for instance along the Baltic coast of Poland, into Hungary, or onto Iceland. Proschwitz (2010) has suggested that in Sweden the ameliorating climate is associated with the recent range extension, and the same process could occur in eastern Europe.
In North America, three sites in Colorado and Utah where D. invadens has been found away from garden centres and plantings of annual bedding plants (Table 2) reach temperatures similarly cold as the extremes in Europe: for Fort Collins, winter minimum temperature over last 20 years averaged -22 °C with a minimum of -28 °C; for both Denver and Snowbird these statistics are -22 °C and -27 °C (http://ccc. atmos.colostate.edu/sum_form.html; http://climate.usurf.usu.edu).
At the other extreme, too much heat is probably not a restriction per se, since the species occurs in Egypt, for instance. Meyer and Cowie (2010) proposed that the occasional extremes of a temperate climate preadapt some invasive molluscs to be able to cope with the high temperatures of the tropics. However, heat may be associated with drought, at least seasonally. In such regions the automatic watering systems in use in intensive agriculture and in horticulture (particularly prevalent in prosperous countries) facilitate the survival and spread of D. invadens. Maybe these artificial habitats will provide a route for D. invadens to colonise naturally damp areas within steppe habitats.
In Central and South America, D. invadens occurs within the tropics but the accurately localised records are all from higher altitudes. This may be because higher altitudes tend to be cooler and have different precipitation patterns, or because such climatic differences have encouraged urbanisation or types of agriculture that favour the species. Higher altitudes would seem the likeliest place to search for D. invadens in other areas of the tropics, including Africa and India.
Deroceras invadens is typically associated with disturbed habitats, especially gardens, and is often easiest to find under discarded rubbish. It is one of the few molluscs to occur in the most urban sites, by surviving in the soil of flowerpots (Horsák et al. 2013b). Greenhouses and consistently watered nurseries suit it, and it occurs in some arable fields (e.g. England, N Spain, Switzerland, Australia), but is a significant pest more rarely than D. reticulatum. Although reported to avoid grassland in Britain (Dirzo 1980), D. invadens occupies that habitat in New Zealand and Australia (Barker 1982, Holland et al. 2007). Outside of the probable natural range of the species (Italy), the species has spread to natural habitats not only in western Europe (e.g. Britain, Ireland and France, although not in Germany) but also in Australia, South Africa and the Canary Islands. It seems likely that dispersal into gardens is often rapid, but that the spread into and across agricultural and natural habitats is a slower process and not inevitable. This could explain the positive association with population density even in long colonised countries such as Britain, Ireland and Belgium.
Deroceras invadens can become one of the most frequently encountered slugs, typically about half as frequent as D. reticulatum, but at some sites even commoner (e.g. Manchester gardens, British commercial greenhouses, Frankfurt am Main, Egypt). In the laboratory we have observed that D. invadens can mature a month faster than D. reticulatum, which may give it an advantage when growing seasons are short owing to either climate or agricultural activities.

Geographical range
Deroceras invadens is widespread over most of the western half of Europe. However, there are still areas within this region where it is scarce. In some cases this is probably because of climate (cold in Scandinavia, summer drought in central Spain), in other areas it might merely be because of a lack of time to spread there (e.g. Suffolk, Alsace). The species has yet to spread far in eastern Europe and is still much more thinly spread in Germany and Austria than in Belgium or Great Britain, for instance. Currently the most eastern outdoor records in Europe are the Baltic coast of Sweden, Wrocław in Poland and Ostrava in the Czech Republic; further east, at least in Lithuania, Latvia and the Ukraine, there is sufficient current interest in slugs that D. invadens would probably have been recorded had it become well established (Skujienė 2002, Gural-Sverlova et al. 2009, Rudzīte et al. 2010. The species also appears largely absent from the eastern Mediterranean (except for one record from Crete and some from along the Nile in Egypt in agricultural habitats). The Balkans and Turkey are a centre of diversity of Deroceras, so maybe competition with local species is a factor limiting the spread here. But this diversity also makes it harder for malacologists to spot a new arrival. The absence of D. invadens from Asia is striking, but have malacologists there been looking critically at their alien Deroceras species?
The species has long been known from the Pacific Northwest of America, and also occurs sporadically elsewhere in the USA and Canada. The pronounced scarcity of records in the east compared with the west is untypical of other introduced European slugs. The reason could be that much of the east has an unsuitable climate for D. invadens; perhaps in areas southern enough to for the winters not to be too cold, the summers are too hot and dry (Fig. 4B). But surely there is enough artificial irrigation often to overcome this restriction. Or, supposing that D. invadens has spread to America much later than most other introduced slugs, perhaps its distribution remains more biased towards the original site of introduction. However, the pattern could also be an artefact: our hunch is that D. invadens is at least somewhat more widely spread in the east than currently recognised, but has been frequently misidentified as D. laeve. This confusion had no doubt hidden its occurrence in Central and South America; the records assembled here are sparse but imply that D. invadens is widespread.
Elsewhere in the southern hemisphere, the species has been present, maybe for a long time, in the former British colonies of South Africa, Australia and New Zealand. In this context, the indication from interceptions that it may be present in Kenya is not surprising. It has also colonised a number of remote oceanic islands; the maritime influence on their climates is probably favourable, and perhaps also their depauperate faunas have left a niche vacant.
The range of D. invadens is impressive (Fig. 4), and current records probably still significantly underestimate it. However, it should be recognised that there are a number of other European slugs that are ahead of it in the extent of their non-native range. Besides D. laeve and D. reticulatum, one can list at least Lehmannia valentiana (Müller, 1774), Limacus flavus (Linnaeus, 1758), and Milax gagates (Draparnaud, 1801) (Herbert 2010). Although these species have spread at different rates, the process is continuing, so it remains to be seen whether the ranges that they eventually occupy will differ significantly.
The similar species D. panormitanum s.s. from Sicily and Malta has also been introduced elsewhere, but much more rarely than D. invadens: the only such records are from one site in northern Italy, one in Wales, two adjacent sites on Madeira and an interception from the Azores (Tables 1, 2). Although both species are probably native on Sicily, we never collected them together there (JMCH & HR unpublished), so the species might tend to exclude each other. We did collect them together from the same culvert in northern Italy (Table 2), but two subsequent visits to the site yielded only D. panormitanum s.s.

Spread
Deroceras invadens has been directly observed arriving from abroad on salads, vegetables, flowers, roots, and tiles (Falkner 1979 and USDA records). Such vectors need not be representative of the processes responsible for spread within a country, which surely must often be via garden plants, considering how frequently the species has been found in nurseries and garden centres. At both scales it seems unnecessary to invoke dispersal on bird's feet or plumage, for which there is conspicuously little direct evidence for any slug (Pearce et al. 2012); birds are a parsimonious explanation only perhaps for colonisation of uninhabited islands (e.g. to Inaccessible Island from Tristan da Cunha).
We hoped that our review of the literature would illuminate the rate and pattern of these dispersal processes, but mostly it is hard to be sure that the apparent rate of spread is actually not the spread of awareness that this novel species is worth distinguishing from others. That is particularly a problem with a slug species that requires dissection for reliable identification. In several cases (e.g. Britain, France, New Zealand) the species was probably widespread before anyone was aware of its presence; presumably at a more local level the distribution continued to grow denser, but usually there are no follow-up surveys once someone has claimed the first record. What is really required is an initial survey reporting absences of the species, then comparable repeat surveys of the same places in subsequent years; this has rarely, if ever, been done.
There is nevertheless good evidence of a spread within one or two decades through the Azores and Tenerife. The German data are also probably reliable and representative in suggesting a time scale of one to two decades to extend over a larger country, but it is far from the case that every suitable garden or even district has been colonised within that time. Puzzling gaps in the present distribution elsewhere (e.g. Suffolk in England, Alsace in France) suggest that "filling in" can take decades longer. It is difficult to make quantitative comparisons between species, especially because the delay in spotting a new arrival depends on the ease of recognising the species, but D. invadens probably has spread a little slower than three other terrestrial molluscs that have also invaded much of Europe within the last century, the slugs Boettgerilla pallens Simroth, 1912 and Arion vulgaris (Moquin-Tandon, 1855) and the snail Hygromia cinctella (Draparnaud, 1801) (Reise et al. 2000, Beckmann and Kobialka 2008, Kozłowski and Kozłowski 2011. Quite probably when D. invadens first arrived in England and France, the more local pattern of trade at that time led to a slower spread than has occurred in more recently colonised European countries, but the quality of the data is insufficient to test this. One also expects the spread to be slower in countries with a hostile climate, which might explain the dearth of additional records in Poland, Slovakia and the Czech Republic. Similarly, the rate of spread in Sweden may reflect recent climate amelioration as much as the dispersal process (Proschwitz 2002). Marion Island is a special case, but is revealing in demonstrating the reliance on man for long-distance dispersal within the island, in contrast to the slow initial penetration of the natural habitat by natural means. Here and elsewhere, it is an interesting open question whether the spread accelerated following a period of genetic adaptation to the local environment. The climatic diversity of its non-native range would make D. invadens an appealing subject on which to test whether such adaptations have evolved.
One would expect uniparental reproduction to facilitate colonisation if adventitious human-mediated transport sometimes introduces a single slug at a time. Foltz et al. (1984) observed that self-fertilising slug species had been significantly more successful in colonising eastern North America, with D. invadens one of the species fitting this pattern (outcrossing and absent). Subsequently D. invadens has been found in eastern North America, but paternity studies in our laboratory (Reise et al. unpublished) have shown that the species does sometimes self-fertilise, producing viable offspring, especially in the absence of a partner. Nevertheless, the intra-site polymorphisms that Foltz et al. (1984) observed in populations from the British Isles and France do not suggest that the species' spread through these regions was dependent on a succession of founder events involving self-fertilisation. In contrast, on the remote Marion Island, all 25 slugs sequenced had the same COI haplotype (Lee et al. 2009). In ongoing work, we are comparing patterns of genetic diversity elsewhere in its range.

Introduction
The region comprising the British Channel and southern part of the North Sea, as well as coastal areas of Great Britain, France, Belgium and The Netherlands (also referred to as The Two Seas region) has a long history of trade and travel, and includes important commercial ports such as Southampton, Felixstowe, Le Havre, Antwerp and Rotterdam ("World Shipping Council. Top 50 World Container Ports" ; Enshaei and Mesbahi 2009). These intensive activities across national borders have led to the introduction of numerous exotic animal, plant and other species to this area, both from other European regions and further afield (Holdich and Pöckl 2007;Gherardi et al. 2009;Keller et al. 2009). Invasive species do not know political borders, which is why cooperation and collaboration between countries is key in the fight against devastating and costly nonnative, invasive species. Efficient cross-border communication and knowledge transfer would guarantee that knowledge on the vectors, impacts and control options for nonnative species gained in one country informs decisions on management and control for non-native species in other countries. It can further help to raise the alarm on species that are likely to spread from one country to another, prompting preventive action plans. International cooperation in environmental politics can facilitate development and implementation of sustainable cross-border management practices for non-native species (Essl et al. 2011).
Examples exist where international cooperation has significantly improved the prevention of non-native species' spread. These include the Inter-American Invasive Species Network (IABIN-13N, http://i3n.iabin.net/) that supports the detection and management of invasive alien species in the Americas, and the Trilateral Committee for Wildlife and Ecosystem Conservation and Management (http://www.trilat.org), which addresses environmental challenges common to Canada, United States and Mexico Simpson et al. 2009).
With regard to the Two Seas region, the European-funded 'Interreg Two Seas Programme' promotes cross-border cooperation between Great Britain, France, Belgium and The Netherlands (see http://www.interreg4a-2mers.eu for more information). Within this initiative, the RINSE (Reducing the Impacts of Non-native Species in Europe) project aims to develop cross-border tools to improve the prioritisation and targeting of non-native species. As a minimum requirement, such a regional approach to invasive species' management requires an up-to-date and comprehensive registry of non-native species containing information on the current status of each non-native species in each of the four Two Seas region countries. To facilitate application in reliable horizon-scanning and similar exercises, and allow for meaningful comparisons of inventories between countries and taxa, it is crucial that this database is as comprehensive as possible but not skewed towards particular countries or taxa.
Unfortunately, currently available databases are unsuitable for such a purpose. For example, the number of species listed in freely accessible online databases differs considerably between databases as well as between the four countries of concern. Thus, the primary data portal for non-native species in Europe, DAISIE (Delivering Alien Invasive Species Inventories for Europe; http://www.europe-aliens. org/), features 2,471 non-native species for Belgium, 2,075 of which are terrestrial plants, but only 881 species for the slightly larger Netherlands. In contrast, the Dutch Biodiversity registry (http://www.nederlandsesoorten.nl) lists 925 non-native species, and only 101 invasive non-natives are highlighted by the Belgian information system Harmonia (http://ias.biodiversity.be). The Great Britain Invasive Non-Native Species Secretariat (NNSS) database (http://www.nonnativespecies. org) comprises over 3,000 species. No comparable initiatives exist in France. Such enormous discrepancies between inventories of neighbouring countries that would be expected to host comparable numbers and sets of non-native species are unlikely to be real but probably root in different experts providing the data for different countries and databases.
The present dataset aims at providing a registry of non-native species in the Two Seas region that is comprehensive and not biased towards particular countries or taxa. We achieve this by integrating information from a total of 55 national and international print-and online-sources on the presence of non-native terrestrial, marine and freshwater species in the four Two Seas region countries. In addition, for each taxon, the registry provides information on its taxonomic classification, current distribution and environment inhabited. Potential utilities of this registry include developing national checklists of non-native species, and analysing spatial patterns of distribution of species. Furthermore, the database offers a general point of reference for both scientists and practitioners working on non-natives in the Two Seas region and adjacent countries. Finally, the registry could act as a tool to assess reliability and comprehensiveness of other databases from which data was retrieved. This could be done by, for example, comparing number and identities of non-native species listed by a given source database to those of the present registry.

A. Data set identity
Registry of non-native species in the Two Seas region countries (Great Britain, France, Belgium and the Netherlands)

B. Data set description
The dataset consists of 1 file, containing two worksheets. Worksheet "Registry" contains a 10 × 6,662 matrix of text values, Worksheet "Summary of data" contains three summarising tables in the form of three 2 × 33, 2 × 5 and 8 × 5 matrices of text and numeric values. The file is labelled as Table_RegistryNonNativeSpecies.xls

Identity
We collected information on the status of non-native species in each of the four countries Great Britain, France, Belgium and the Netherlands. The taxonomic affiliation and environment inhabited by each species is also provided.

Originators
The project was conducted within and on behalf of the European Union funded RINSE (Reducing the Impacts of Non-native Species in Europe) Project. Methodology was developed by the three authors Alexandra Zieritz, Belinda Gallardo and David C. Aldridge. Data collection was done by Alexandra Zieritz and Belinda Gallardo.

Period of study
Data was collected from 01/06/2012 to 27/02/2014. Collected data correspond to contemporary species records.

Objectives
The primary objective of the present work was to compile a registry of non-native species present in the four countries comprising the Two Seas region, i.e. the English Channel and the Southern part of the North Sea. The registry will serve as a basis for developing effective, cross-boundary strategies to manage and control non-native species, which can have severe ecological and economic impacts. It can further be used as a general reference for both scientists and practitioners, as well as a tool to assess reliability and comprehensiveness of other well-known databases such as the DAISIE portal.

Source of funding
INTERREG IVA 2-Seas Programme. Funded by the European Regional Development Fund (ERDF). Project: RINSE, Reducing the Impacts of Non-Native species in Europe. Work Package 1 subproject: Targeting and Prioritisation for Non-Native species into the RINSE area.

Summary of the RINSE project
RINSE (Reducing the Impacts of Non-native Species in Europe) is a European Project which investigates best strategies of managing non-native species (NS) across the Two Seas Programme area. The project specifically aims to i) develop cross-border tools to improve prioritisation and targeting of NS, so that scarce resources can be directed towards the species and sites of greatest concern, ii) enhance the capacity to address NS within a range of target stakeholders, and iii) develop new approaches and best practices for the management of NS, by delivering field trials and demonstration projects. RINSE works across borders to share best practice and adopt strategic approaches to tackle the threats posed by non-native species (NS).

B. Specific subproject description 1. Site description. a. Site type
The region includes terrestrial, marine and freshwater habitats.

b. Geography Location
Countries comprising the Two Seas Programme area (i.e. the English Channel and the Southern part of the North Sea): Great Britain, France, Belgium and The Netherlands.

c. Habitat
The region includes terrestrial, marine and freshwater habitats.

d. Geology, landform
The region includes various geological types, ranging from Pre-Cambrian, to Carboniferous, Cretaceous and Tertiary rocks.

e. Watersheds, hydrology
The main river systems in the area include the Thames, Loire, Seine, Meuse and Rhine.

f. Climate
Climatic conditions in the study area range from Mediterranean (i.e. Southern France) to temperate.

Experimental or sampling design a. Design characteristics
Basic data were collected by systematic review of 36 web-and print-based sources over an eight-month period (see Table 1). Additional systematic scanning of three scientific journals, i.e. Neobiota, Aquatic Invasions and BioInvasions Records, recovered 19 additional relevant publications from which information was included in the registry.

b. Data collection period, frequency, etc.
Basic data collection period was 01/06/2012 to 17/01/2013. Additional systematic scanning of the three journals Neobiota, Aquatic Invasions and BioInvasions Records was performed in January and February 2014.

Research methods
A number of online and print data sources were used to obtain information on nonnative species present in the four Two Seas region countries. In total, the basic data were gathered from 36 sources, including the 12 listed in detail in Table 1 used for all animal phyla and plant divisions, and additional references for particular groups of organisms (see references). Selection of the databases included in this work was done with the help of consulting experts within the European RINSE (Reducing the Impacts of Non-native Species in Europe) project. Particular care was thereby taken to avoid and counteract any bias towards particular countries. For example, the lack of a national database on non-native species in France was targeted by inclusion of an additional 11 grey-literature sources from France, which we obtained through our local RINSE partners ("Le Conservatoire Botanique National de Bailleul. Liste des plantes exotiques considérées comme envahissantes en Picardie"; Agence de l'eau Artois Picardie and Conservatoire   Other abbreviations: na, not applicable because no data on presence of species within the respective phylum/division in the four countries were available from this source; N, source not used with regard to respective phylum/division; Y, all taxa of respective phylum that this source lists to be present in one or more of the four countries were included; Y ex T, all taxa except terrestrial ones of respective phylum that this source lists to be present in one or more of the four countries were included.
Additional 17 sources were used for following groups: • Angiospermae: Agence de l'eau Artois Picardie and Conservatoire Botanique National de Bailleul (2005) Furthermore, all volumes of the three journals Neobiota (vol. 9-20), Aquatic Invasions (vol. 1-8) and BioInvasions Records (vol. 1-2 and vol 3 in press articles) available by February 2014 were thoroughly and systematically scanned for relevant studies that potentially provided further information on species' presence in the RINSE countries. This was done by reading the titles and, in case that indicated potential relevance to our database, reading the abstract and complete manuscript. In total, the following 19 studies were thereby included in the registry: Copp et al. (2006), Kerckhof et al. (2007), Sjøtun et al. (2008), Kai and Soes (2009) Lavesque et al. (2013), Minchin et al. (2013), Pinder et al. (2013), Scalone and Rabet (2013), and Soors et al. (2013).
Finally, apart from collecting basic data on non-native species present in at least one of the four Two Seas region countries, additional presence in the other three countries of concern was checked using the following seven geographic distribution gateways: "GBIF -Global Biodiversity Information Facility. GBIF Data Portal", Hopkins (2012), "Intergovernmental Oceanographic Commission of UNESCO. The Ocean Biogeographic Information System OBIS", "Muséum national d'Histoire naturelle. INPN Inventaire national du Patrimoine Naturel", "NBN -National Biodiversity Network. National Biodiversity Network's Gateway", "NLBIF -Netherlands Biodiversity Information Facility. Data portal of the Dutch national node of the Global Biodiversity Information Facility (GBIF)", and Verloove (2006). This was done for all taxa except those terrestrial Angiospermae and Pinophyta that were listed as present by the DAISIE portal but not by any of the other databases consulted. In the registry, these species are indicated by the phrase "data based solely on DAISIE portal" in the final column (headed "Notes").
After compilation of the database was completed, we checked for errors through the process of blind repetition of data-compilation for 1% of the dataset (i.e. for 34 species or 136 data points (34 species × 4 countries)). This revealed an error rate of 0.007% (i.e. 1 of 136 data points was incorrect).
The present database will be sustained in the future by periodically conducting a systematic literature review on new invasions in the four countries. This could be done, for example, by a Web of Science or Google Scholar search using keyword combinations such as "non-native OR exotic OR invasive AND Britain OR UK OR Netherlands OR France OR Belgium", and/or a systematic scanning of the most relevant journals such as Nebiota, Aquatic Invasions and BioInvasions Records.

Size
The size of the file is 389 KB. The table lists 6,661 species, subspecies and hybrids. In total and including headers, the "Registry" worksheet therefore contains 66,620 cells. The "Summary of data" worksheet contains 92 cells.

Format and storage mode
The file type is MS Excel®. No compression scheme was employed.

Header information
A single header row includes the species' phylogenetic classification (i.e. four headers: phylum/division, class, genus and species name), status in the four countries investigated (i.e. present, native, extinct or not confirmed), environment (i.e. terrestrial, freshwater, marine, freshwater+terrestrial, marine+freshwater or terrestrial+marine), and Notes.

Alphanumeric attributes
Alphabetic character fields. Great Britain Status of species in Great Britain N/A Character extinct -non-native species was present in the wild in GB/France/Belgium/Netherlands at some time but is no longer present in the respective country native -species native to GB/France/Belgium/ Netherlands not confirmed -presence of non-native species not confirmed for GB/France/Belgium/ Netherlands present -non-native species has been recorded in the wild in GB/France/Belgium/Netherlands and is likely to exist there at this time present/extinct -non-native species listed as "present" by one source but as "extinct" by another source present/native -non-native species listed as "present" by one source but as "native" by another source Taxonomic uncertainty in pest risks or modelling artefacts? Implications for biosecurity policy and practice

Introduction
In a recent issue of Australasian Plant Pathology,  presented modelling results on the high profile plant pathogen Puccinia psidii, which causes a major disease of plants within the Myrtaceae family. The disease is commonly referred to as guava or eucalyptus rust. This case study was used to explore the effects of taxonomic uncertainty on pest risk assessments for the Australian continent, and it was concluded that the estimation of pest risk can be highly sensitive to different taxonomic delimitations of the organism under investigation. The model variations either considered Uredo rangelii (commonly referred to as myrtle rust) a different species to P. psidii sensu stricto (s.s.), or a minor morphological variant within the broader species complex P. psidii sensu lato (s.l.) Carnegie and Cooper 2012). Our aims with this comment are to contribute constructively to modelling best practice for pest risk assessments, and to highlight our concerns with using the results presented in  as a basis for estimating geographical pest risks from P. psidii. We first highlight what we consider to be a number of methodological limitations in the models of , and give context to the attribution of causation that follows. Considering the importance of, and interest in this pathogen in Australia and worldwide, we also discuss the implications of the results and conclusions presented in  for biosecurity management.

Taxonomic historical context
For at least one hundred years, several names had been applied to different populations of the rust fungus found on Myrtaceae in South America, now all regarded as P. psidii s.l. About thirty years ago, it was found that two different types of urediniospores are present in voucher specimens of P. psidii (Walker 1983). In some cases, the urediniospores are echinulate all over their surface; in the others, there is a smooth patch (called a tonsure) free of spines on the spores. These observations led to the proposition that P. psidii may be a complex composed of different morphological, and perhaps physiological, forms. In 2006,  named the entity within P. psidii with tonsured urediniospores Uredo rangelii, based only on two herbarium specimens that did not have teliospores. This new species determination is controversial, since the presence of a tonsure on urediniospores in some populations (with or without teliospores present) is the only distinct morphological variant found within P. psidii s.l. It is widely believed that more research on the morphology, host range and molecular characteristics of different populations is necessary before it can be decided if the variant with tonsured urediniospores and other putative variants should be considered as different varieties or distinct species. Until these questions are resolved, the rust on Myrtaceae can be referred to as P. psidii s.l., or the P. psidii complex.

Methodological issues and concerns
The modelling system (MaxEnt; ) that  used to support the sensitivity analysis of pest risk maps is a commonly used correlative model. MaxEnt fits maximum entropy statistical models to relate geographical distribution points to spatial covariates. There are three issues worth elaborating regarding the development of the models presented in , and consequently the interpretations drawn from them.

Unstable covariate importance rankings
The covariate importance rankings in the MaxEnt models of  varied substantially and inconsistently between taxonomic treatments. In order to assess the stability of the covariate rankings we applied Kendall's rank discordance test (Zar 1984) to the MaxEnt permutation importance scores for covariates of the  models (we believe this to be a more relevant score than percent importance; Table 3 in . The value of 0.21 for Kendall's coefficient of concordance, W, indicates that the ranking of the weighting of the covariates is highly discordant across the models. Instead of fitting modified response functions to the single set of covariates, MaxEnt fitted functions to different sets of covariates or weighted them significantly differently. Rodda et al. (2011) commented on the instability of a MaxEnt model of the Indian Python described by Pyron et al (2008). Rodda et al. (2011) criticised the "data dredging" practice employed in MaxEnt using the default settings as a factor contributing to this form of model instability, leading to the virtually certain generation of "spurious results". Even using a modest number of covariates as  did, model instability can arise if the covariates are highly correlated or if the model is being challenged to discriminate distribution patterns using covariates poorly related to the ecological processes giving rise to the observed species distribution pattern. An instability in the covariate importance rankings in this type of analysis provides an indication that the model may be unsound, particularly where they are in response to small changes in the training dataset.

Choice of background layers
MaxEnt requires the modeller to specify pseudo-absence data in the form of a defined background layer that spans the known positives, and to a greater or lesser degree, regions in which there are no presence records (Barbet-Massin et al. 2012). The MaxEnt algorithm then fits multivariate functions to selected covariates to try to explain the observed pattern of presence and (pseudo-)absence within the geographical range of the defined background. Although formally defined as the "available environment in the region of study" (Phillips et al. 2009) or the "full environmental range of the species and exclude[ing] areas that definitely have not been searched" , in practice the creation of background layers is an arbitrary process. There is a growing appreciation of the sensitivity of modelling results to the definition of the background (VanDerWal et al. 2009;Rodda et al. 2011;Webber et al. 2011), but yet no universally satisfactory method for doing so.
In framing the modelling treatments,  presented four distribution data subsets from within the known distribution records of P. psidii s.l. (referred to as Puccinia_94).  then prepared different backgrounds for the subsets identified as U. rangelii and those of P. psidii (s.s. or s.l.), though clear definitions of these backgrounds were not presented in either the main document or the online supporting materials. Irrespective of this oversight, from an experimental design perspective, we believe that the various taxonomic treatments subsequently modelled in  are potentially confounded by the interplay between model background definition, the covariate space of distribution data, and the subsequent response curve characteristics and form.
It may be argued that the taxonomic in silico experimental treatments in Elith et al. (2013) are inherently compound treatments when considered in MaxEnt, and that it is unrealistic or inappropriate to change the input data points without also changing the background definition. Whilst we have sympathy with this argument, modifying the arbitrary background between distribution data subsets interferes with attribution of causality when considering model results. This methodological issue arises with all correlative models that require pseudo-absence inputs, questioning their suitability for representing or explaining the real world differences in pest threat from the different taxonomic delimitation treatments. Figure 2 of  indicates that the xeric and alpine regions of Australia are 'suitable' for U. rangelii, which, on the balance of evidence seems highly improbable. We suspect that both the cold and dry tolerance limits in the MaxEnt Uredo models are strongly influenced by a single misleading location record in north-western Argentina that was included in the training dataset. This record appears to be a coarsely geocoded point location representing the centroid of Tucuman, a small province that spans a large altitudinal (and temperature and rainfall) range at the base of the Andes. We think it is likely that the centroid falls in colder and drier conditions in the gridded climatology than where the specimen was observed. Similar improbable projections for the rest of the world (Elith et al. 2013, Online Appendix 3) are also not discussed, despite their potential implications for pest risk management outside Australia.

Covariate values and extrapolation
MaxEnt requires the modeller to specify how the model should fit covariate response functions beyond the range of conditions experienced by the training dataset (Fig 8 in Webber et al. 2011;Webber et al. 2012 Fig. 3, Online Appendix 5), it is clear that the "clamping" option was chosen. This choice means that the fitted response value for the most extreme covariate value was held constant when the model was extrapolated beyond the training data. Biologically, this default option in MaxEnt defies the Law of Tolerance (reviewed in Shelford 1963). Furthermore, no threshold was set to divide modelled suitability values into 'suitable' and 'unsuitable' regions. To explore the implications of these choices and the form of the response curves more generally, we constructed a conceptual sketch of how projections might have been influenced by the modelled relationships (Fig. 1). In doing so, we recognise that the MaxEnt 'Explain' tool is also particularly informative for this type of interrogation, should all the input data be available.
Developers of MaxEnt have called repeatedly for "extreme care" when extrapolating to novel climates (e.g. , going as far as stating that extrapolation is "inherently risky" . As such, like many correlative species distribution models, it has been shown to be poorly suited to addressing questions concerning novel climates, such as encountered in biosecurity and climate Inland areas of the Ured_27 model are modelled unfeasibly as having areas of relatively high suitability in excessively dry climates. In c and d, the fitted quadratic response functions asymptote, rather than intersect the origin, leading to biologically unfeasible modelled climate suitability in excessively cold areas ). change applications (e.g., Kriticos and Randall 2001;Sutherst and Bourne 2009;Rodda et al. 2011;Webber et al. 2011). MESS maps (sensu ) and the ExDet tool (Mesgaran et al. 2014) provide an informative way to visualise regions where the model is extrapolating, and therefore, where any model interpretations should be done extremely cautiously.  state "we have evidence that the models do not need to extrapolate in our regions of interest". However, the MESS maps presented in the paper (their Online Appendix 4) are based on the extent of the backgrounds (rather than the extent of the known presence points), only identify model extrapolation due to covariate range, (ignoring changes in the relationships between covariates), and use an ambiguous colour scheme to depict extrapolation in the geographical regions where biologically implausible areas are modelled as suitable within the "area of interest" (Australia; Fig 1). These choices and the form of the fitted model response curves result in potentially unsuitable environments being modelled as suitable beyond the putative physiological limits of the species concerned (i.e. commission errors).  states "we use the Australian incursion of myrtle/guava rust as an example, not to argue which taxonomic interpretation, data set or model is correct, but to highlight the impact of taxonomic belief on modelled predictions". However, there is no denying that the results presented will attract the attention of many biosecurity managers, both within Australia and worldwide. These results, therefore, fall within the context of a discussion about the degree of risks Australia faces from a recently discovered, high profile, non-native pathogen whose taxonomic status is uncertain.

Concerns with possible impacts on policies
Maps are potent communication devices, and two aspects of the way they have been presented in  could have significant impacts on how they are interpreted by biosecurity practitioners. Firstly, the white regions in the maps in Figs. 2 and 4 in  may imply that a modelled suitability threshold of 0.05 was chosen, particularly as white was also used for the 'no data' class. Discussions with the authors confirmed that this is not the case, and the classification threshold was an arbitrary cartographic choice; hence, there were no explicit assumptions made about where biosecurity managers should 'stop worrying' about the threat of establishment of P. psidii s.l. These models, therefore, provide limited information for delimiting areas at risk of establishment or invasion (compare Fig. 1 of this paper with Fig. 4 of . Secondly, a reasonable interpretation of these maps appears to be that the extent of the geographical area of concern, at all levels of modelled suitability, is likely to be greater from U. rangelii than P. psidii s.l., despite U. rangelii distribution records having a narrower geographical range. Indeed, this is the interpretation intended by Elith et al. (2013:49): "Recognition of P. psidii sensu lato (Puccinia_94) would lead managers to place lower priority on surveillance and containment in Western Australia, and to increase the focus of activities in Australia's northern and eastern neighbours (e.g. New Caledonia). Recognition of U. rangelii (e.g. Uredo_27) would lead managers to increase the priority for these activities in New Zealand, Tasmania and Western Australia (Fig. 2).". According to both set theory and ecological reasoning this result and conclusion are implausible. In ecological terms, based on the critical assumptions underlying correlative modelling methods and niche theory, the broader the range of environmental tolerances encompassed by an organism in its native range, the broader the range of conditions we might suppose it is at least capable of inhabiting in an introduced range. That is, if Area small and Area large refer to the environmental space occupied by taxa with a smaller or larger environmental envelope respectively, the corresponding environmental space projected to be suitable (potential niche breadth, Habitat small and Habitat large ) should conform to the same inequality i.e., because Area small ⊂ Area large then Habitat small ⊂ Habitat large . The results presented in  indicate the opposite, and are most surely a modelling artefact, due to the reduced information in the smaller datasets in relation to the background definition. Importantly though, the manner in which the MaxEnt model assigns probabilities to cells means that the map classes in the different maps in Fig. 2 and Fig. 4 in  do not, strictly speaking, represent the same level of modelled suitability. That is, the values of modelled suitability represented in these maps cannot be compared directly between maps. This limitation is a subtlety which is not conveyed in the figure captions or body text of , and we contend is not something that would be generally appreciated within the risk assessment community. Taken together, policies based on the prima facie results of the comparative modelling presented in  would be in the wrong direction in terms of the perceived geographical extent of the threats.
From ecological theory, we expect that closely-related species (such as U. rangelii and P. psidii s.s., if they are eventually confirmed as being different species) may competitively exclude each other from otherwise suitable habitat (Hardin 1960;Mitchell and Power 2003), though for some species including an overlapping hybrid zone. This principle underlies the Enemy Release Hypothesis (Keane and Crawley 2002), which describes the direction of expected niche changes when organisms are freed of the effects of their natural enemies, including that of competitors. Evidence of this may be found in comparisons of habitat suitability from non-native and native ranges (Kriticos and Randall 2001). It also underpins cautions to pest risk analysts to consider interactions with other species (Davis et al. 1998) and the effects of land use in modifying the apparent climatic response (Bourdôt et al. 2013).
A corollary of this theory is that the inferred differences in climatic preferences based on native range sampling relate only to the realised niche (Hutchinson 1957;Soberón 2007); the fundamental niche of closely-related sympatric species may overlap strongly. Hence, for such species, taxonomic hair-splitting when selecting presence records for modelling is likely to produce conservatively-biased results. This method may be attractive for conservation-oriented modelling; however, the costs associated with errors regarding biosecurity risks are asymmetric. That is, the consequences of underestimating the potential range of an invasive alien species are that it could become established, creating perpetual unwanted pest impacts. The consequences of overestimating the risk are unwarranted resource allocation to biosecurity exclusion efforts. Ideally, both types of errors would be minimised when informing biosecurity policies, but overestimating the risks is generally to be preferred to underestimating them. However, the overestimates should result from ecologically robust modelling projections, rather than from modelling artefacts.
Based on their modelling,  argue that significant resources should be devoted to resolving the taxonomic imbroglio of this rust fungus. We agree that it is critical generally for ecological modellers to attempt to clarify the taxonomy of the organism being modelled. This applies equally to geographical distribution data (Kriticos et al. 2003;Dupin et al. 2011;Baker et al. 2012;Bourdôt et al. 2013) as it does to other biological data about the taxon of concern. However, from a pragmatic biosecurity management perspective, attempting to narrow the perceived pest risks from P. psidii s.l. to those from U. rangelii faces two distinct challenges. Firstly, our ability to readily detect and distinguish samples of P. psidii s.l. and U. rangelii are impractical, and secondly, the management responses would be no different depending on which of these taxa are posing an invasion threat. We contend that in cases of taxonomic uncertainty, robust decision making in the biosecurity arena (sensu Simon 1991) is therefore best informed by assuming the broader taxonomic delimitation of P. psidii s.l., as has been done in Australia and elsewhere to date for surveillance and monitoring efforts. The expenditure of significant resources on taxonomic differentiation for pest risk assessments may result in considerable time and resource savings for some systems such as the Tephritids (Schutze et al. 2012), but for P. psidii s.l. it would seem of little practical value within a pest risk management context.
In order to better gauge the biosecurity implications of taxonomic uncertainty (avoiding the pitfalls in the MaxEnt example discussed here), we may be better off using true presence-only correlative models such as BIOCLIM/ANUCLIM (Booth et al. 2014) or CLIMEX Regional Match Climates (Kriticos 2012) that are likely to respond to changes in taxonomy in an ecologically conformal manner. Further, because all correlative bioclimatic models rely upon geographical data solely to infer climate suitability, they are inherently naïve with respect to competitive exclusion and the effects of enemy release. Their ability to extrapolate into novel climates is unreliable because they are not constrained to fit biologically meaningful covariate response functions (Austin 1987), and may over-fit to biased samples (Kriticos and Randall 2001;Rodda et al. 2011;Webber et al. 2011). Another option is to develop more mechanistic models using packages such as CLIMEX Compare Locations (Sutherst et al. 2007b) or NicheMapper (Kearney and Porter 2009), that have been designed specifically for producing ecologically plausible results when projected to novel climates (Sutherst and Maywald 1985;Magarey et al. 2007;Sutherst et al. 2007b;Kearney and Porter 2009;Sutherst and Bourne 2009;Webber et al. 2011;Sutherst 2013). Nonetheless, even armed with models well-suited to the task, it may be difficult or impossible to calibrate models of closely-related taxa sufficiently well to understand accurately the invasion risks they each pose (Sutherst et al. 2007a).

Concluding remarks
Pest taxonomy is a subject that continues to challenge biosecurity agencies (deWaard et al. 2010;Collins et al. 2012;San Jose et al. 2013). Pest risk modelling clearly has a potential role alongside molecular biology in unravelling such taxonomic puzzles, testing taxonomic hypotheses against biogeographical data . However, to earn and maintain this role, the modelling should be reliable, and founded solidly in ecology. In the case of P. psidii s.l., it is our opinion that the modelled taxonomic sensitivity of pest risks presented in  reflect significant modelling artefacts.
The specific issues we raise in this paper fall within the context of a relatively immature, rapidly evolving field of science, where methods are being adapted, developed and tested at such a rate that a consensus view of best practices has yet to emerge. As has been so carefully emphasised in the past by the authors of both this paper and , modellers need to match the question being addressed with the most appropriate techniques available. Whilst this need applies generally, it is particularly pertinent for studies involving biological invasions or climate change where modellers are challenged with unstable range dynamics and projecting results for novel environments (Sutherst and Bourne 2009;Webber et al. 2011). The complexity of biological invasions and the related risk management questions mean that this matching process demands a significant level of modelling experience and expertise.
Some recent developments in computing technologies have been focused on making ecological modelling tools accessible to the masses (e.g., Graham et al. 2010). Even casual scans of ecological modelling discussion-lists reveal the alarming frequency with which scientists with little or no prior experience or training in ecological modelling grasp these techniques and apply them as a means of 'rounding out their studies'. The dire consequences are apparent in the proliferation of poorly-founded and ecologically-implausible models appearing in the recent species distribution and niche modelling literature. Clearly, making it easier to generate models does not ensure that the models will be meaningful or fit-for-purpose. Modelling species' current and potential distributions is a complex, difficult task, and there are few formal opportunities for learning appropriate modelling skills.
For end users there is a need to become more familiar with how various modelling choices affect the meaning and utility of the model results. Moreover, modellers have a responsibility to foster an effective understanding of these issues amongst biosecurity risk managers. For example, it is not possible to look at a map of a modelled species distribution, and to know instinctively what it means in terms of pest risks. There are many different ways in which the risk modelling problem can be framed, and the meaning of the model results changes accordingly.
There is clearly still much work to be done in this space, and we will need contributions from both the ecological modelling and biosecurity communities to achieve our goals of advancing best practice for pest risk modelling. these groupings could be debated and that future evidence might prove them wrong. Our aim ... is not to argue that the specific choices are indisputable, but to create five distinct datasets with reasoning behind each, and to use these to model and predict". In other words, we were not trying to model these entities definitively, but to explore the effect on predictions of different perspectives on the group's taxonomy. We used the modelling software Maxent Dudik 2008) and 7 covariates selected for their likely ecological relevance to model these five datasets. Kriticos et al. (2014) note that the covariate rankings in the five models vary, which they interpret as 'unstable'. They state "instability in the covariate importance rankings in this type of analysis provides an indication that the model may be unsound". In Maxent, variable importance is estimated in two ways -one during model building and the other on permutation tests on the final model. Kriticos et al. (2014) focus on the second and expect similar covariate rankings across all 5 datasets. We agree that repeated random samples of the full distribution of a species are likely to lead to similar covariate rankings, provided there are enough samples to reliably model the species and provided the entity is in fact a single species. We emphasise, though, that we were not taking repeated random samples. Many circumstances could lead to different covariate rankings. These include different datasets representing different species or subspecies, a random sample that happened to be biased to one part of covariate space, or highly correlated variables. In the latter case, different covariate rankings may lead to very similar predictions because covariates are largely interchangeable. There is no reason to impose an a priori expectation that covariate rankings should be similar across our 5 datasets, not the least because we were exploring the possibility that some subsets display different environmental constraints because they represent different taxa. Kriticos et al.'s arguments are pre-conditioned on the assumption that we are dealing with one species. We held no such presumption. Kriticos et al. (2014) used Rodda et al. (2011) as a support for their argument regarding the stability of covariate rankings. We fail to see the connection with our work. Rodda et al. (2011) discussed rote use of Maxent with default settings and the commonly available suite of 19 Worldclim variables. They specifically stated that their remarks did not apply to "execution of Maxent with different (i.e. customised) settings". We did not use default settings and 19 variables, and even explored the effects of our choices on the modelled outcomes. Kriticos et al. (2014) note that the choice of background affects model output. They see this -together with the other issues they address -as such a problem that they suggest instead using presence-only methods that do not require background points. It is well established that choice of background affects model outcome (e.g. . In , we described our approach to selecting the background and mapped examples of the background extent. We used a strategy consistent with the approach of an informed user   Fig 1,  grey areas). We did not mention the extent of mapping in the figure legend. We agree that this omission was an oversight, and that generally authors should be specific about their choice of background. We tested the effect of varying the background on model results during model construction and found that it was insubstantial. We agree that it would be better to include such results in appendices.
However, the solution to this issue is not to do as Kriticos et al. (2014) suggest and resort to presence-only methods that do not require background points. Methods that do not take a background sample are ignorant of environmental conditions in the region in which the model is trained. This confounds the frequency of environments at occupied locations with the frequency of environments in the region, an issue that is difficult to overcome and that leads to decreased predictive performance, at least in a number of tested equilibrium situations (Elith et al. 2006). We are left with the problem that choice of background is to some extent subjective, and we suggest that good practice includes exploring the effect of choices on modeled results. Kriticos et al. (2014) question covariate values and extrapolation, in particular the process of clamping predictions in novel environmental domains. We agree that clamping will affect predictions in novel space. That is the intent. Clamping is the default choice and is quite commonly used, including by one of the authors of Kriticos et al. (2014) in . It is a sensible default because it ensures that predictions outside the sampled range are at least consistent with those made at the most similar sampled environment.
Nevertheless, we agree that predictions in novel environmental space should be treated very cautiously. In , our interpretation and interest was focused on areas of relatively high predictions in Australasia and none of these were in novel space. To guard against unintended uses of our results, we could have masked out all regions with any amount of extrapolation, to make it clearer that (1) we were not focusing on any of these areas, and (2) we do not trust predictions in extrapolated areas. Instead, we took the more conventional path of including numerous messages throughout the manuscript and its appendices that showed the reader that extrapolations are inherently uncertain and should be treated cautiously.
Rather than dealing with every issue in the Kriticos et al. (2014) commentary, we now address a remaining important point regarding areas and extents. As Kriticos et al. (2014) note, Maxent (and other methods for modeling presence-background data) outputs relative probabilities or relative intensities. We agree that it would have been useful in  to remind readers of this point specifically, in case they were inclined to misinterpret them as probabilities of occurrence. Nevertheless, the treatment in  was consistent because it focused on relatively suitable locations, those with the highest relative predictions for each taxon. We have discussed this with Kriticos et al. in person, so it is disappointing to find that they persist in stating that  focused on extent.
Regarding "Predicted area", Kriticos et al. (2014) appeal to "both set theory and ecological reasoning" when discussing "Area" and "Habitat". They then apply their reasoning -largely posed in environmental space -to our results, which they interpret in geographic space. They state "the broader the range of environmental tolerances encompassed by an organism in its native range, the broader the range of conditions we might suppose it is at least capable of inhabiting in an introduced range". The discussion in Kriticos et al. (2014) on this point seems at least unclear if not wrong. In fact, range in geographic space depends on the environments available in each region, and that the tolerances of taxa, even closely related ones, are not necessarily nested in environmental space.
Consider this situation: species A and B exist as native species in a certain region. A might be widespread, and B more narrowly distributed. Only a subset of environments that species A occupies in its native range might exist in a new region, whereas all the environments occupied by species B might exist -in fact, in the new region more environments suitable for B might exist than were available in the native range. Thus it is possible (though completely dependent on the relationships between environmental and geographic space) that species A may tolerate a wider range of environmental conditions than B in their native ranges, but species B may have a wider geographic range in a new region. Thinking further about relationships between environmental and geographic space (see the excellent discussion of these issues by Colwell and Rangel 2009): the same suitable environments might be repeated many times in geographic space (the biotope) in a new region, implying that geographic areas are difficult to predict conceptually from environmental thinking. We will not take this argument further since it is difficult to discuss fully in a short reply. Our main point is that whilst Kriticos et al. (2014) might want to interpret our results in terms of areas, we did not make inferences about areas, and their arguments regarding ranges in environmental and geographic space are neither necessary nor sufficient.
Predicted area is not necessarily identifiable from presence-only data. Some authors use 'thresholding' (setting all values above some predicted value to 1 and all below to zero). However, this is not a remedy, since this merely serves to decrease the amount of information available. In  we discussed the effect of small sample sizes (generally predictions will be less well differentiated with small samples) and interpreted the results accordingly.
Lastly, Kriticos et al. (2014) raise a general theme that the results in  will be (mis-)used for other ends. This of course is possible.  focused on the influence of taxonomic uncertainty on the predicted location of the most suitable sites for a set of putative taxa. Those wanting to use the results for other purposes should do so thoughtfully. Our hope is that people wanting to use the predictions in other contexts would either know about or inform themselves about the modelling methods sufficiently that they can make competent interpretations or contact the authors of the original work and ask about their interpretation.
In conclusion, while we appreciate the motivation of Kriticos et al. (2014) was to improve modelling practice and we agree with several of the points they raise about clarity and completeness of descriptions, we find that their main methodological complaints are based on misunderstanding of the technical details of the methods themselves. Our concern and motivation in responding is that their advice is broadly misguided. We agree with some of their concerns regarding the limitations of using correlative methods to model distributions of invasive species (see Elith 2014 for a detailed discussion). The difficulties in predicting species distributions in novel environments remain open and important questions.