Predicting current and future global distribution of invasive Ligustrum lucidum W.T. Aiton: Assessing emerging risks to biodiversity hotspots

Biological invasions represent one of the main anthropogenic drivers of global change with a substantial impact on biodiversity. This impact can be particularly acute in biodiversity hotspots. Ligustrum lucidum W.T. Aiton is a tree species native of China that, from as early as the eighteenth century, has been introduced broadly around the globe, becoming a serious invasive species. We aim to predict L. lucidum's current and future potential distributions at a global scale and assess the degree to which biodiversity hotspots are at risk of invasion.


| INTRODUC TI ON
Invasive alien species (IAS) are one of the five direct drivers of global biodiversity loss (IPBES, 2019). Predicting their expansion under different scenarios is paramount for designing and implementing active surveillance and management actions (Essl et al., 2011;Roy et al., 2014). Human agency plays a critical role in their distribution (Vaz, 2018). Nevertheless, in the case of invasive plants, climate, soil properties and dispersal capacity are three of the most important determinants and constraints to their distribution (Dupin & Smith, 2019;Figueiredo et al., 2017;Liao et al., 2020). Some invasive plants can be limited to growing in areas with similar climatic conditions as in their native range (Peterson, 2003;Petitpierre et al., 2012;Wiens & Grahams, 2005).
However, during an invasion process, other species may additionally expand into regions exposed to different climate conditions than they experience in their native range (Atwater et al., 2018;Broennimann et al., 2007;Gallagher et al., 2010;Goncalves et al., 2014). This characteristic of many invasive plants provides a particular challenge in predicting suitable regions for the species under current or future climate conditions. Glossy privet (Ligustrum lucidum W.T. Aiton, Oleaceae) is a highly invasive tree species present on all continents except Antarctica (Fernandez et al., 2020). The species' original distributional range is thought to have been constrained to the centre-west of China, including Yunnan, Guizhou, Sichuan and Hunan provinces (De Juana, 2009). However, due to dispersal events and long-standing and diverse human uses, the species occurs currently among a broader range of edaphic and climatic conditions along the southern half of China (Chang & Miao, 1986; Editorial Committee of Flora of China CAS, 1999;Wilcox, 2000). Outside its native area, L. lucidum has been introduced around the World, firstly into the United Kingdom by European horticulturalists in 1794, and since then, it has been widely used as an ornamental in Europe (Chittenden, 1951).
The species' introduction was commonly associated with ornamental purposes (CABI, 2020;Hummel et al., 2014;Johnson, 2009;Rodrigues et al., 2015). However, more complex historical and socioeconomic causes, as human migrations and/or trade between the donor area of propagules and the receiving areas, could also have played an important role in L. lucidum's distribution (Madelón et al., 2021;Montti et al., 2017).
To date, scientific research on L. lucidum has mainly focused on its biology and functional traits (Aguirre-Acosta et al., 2014;Aragón et al., 2014;Montti et al., 2016). There has been a particular focus on the species' negative impacts, including a severe decrease in native biodiversity (Ayup et al., 2014;Ceballos et al., 2015;Ferreras et al., 2015;Furey et al., 2014;Malizia et al., 2017) and changes on environmental variables such as reduction in soil humidity and light availability affecting nutrient cycling Whitworth-Hulse, 2018;Zamora Nazca et al., 2014). Further work has investigated the species' role in fostering human allergic diseases (Robledo-Retana et al., 2020). Most importantly, in terms of projecting regions that are likely to be suitable for the species, it was recently demonstrated that invasive populations have expanded into warmer and wetter climatic conditions than native ones (Dreyer et al., 2019).
In recent years, there has been growing recognition of the particular risks that invasive species may pose for biodiversity hotspots and protected areas (Bellard et al., 2014;Li et al., 2016;Liu et al., 2020).
Thirty-five biodiversity hotspots have been defined based upon the high levels of species richness and endemism that they possess (Mittermeier et al., 2004). Together, these areas contain over 75% of the World's vertebrate species and over 50% of the World's plant species. Thus, these biodiversity hotspots are a clear priority for conservation actions, and there are indications from recent work that these regions may be disproportionately at risk from invasive species (Li et al., 2016). Given the dramatic impact that L. lucidum can have on the ecosystems it invades, including dense monodominant stands, it is important to assess the possibility it has to invade these biodiversity hotspots under current and future projected conditions. Because L. lucidum is an invasive species, promoted mainly as an ornamental plant, identifying areas that could be environmentally suitable for the species is a priority for developing early detection networks, control strategies and trade regulations. However, studies on global distribution, habitat suitability or how future climate change can impact the species' potential distribution are still lacking. Species distribution models (SDMs), which aim to predict environmental suitability by linking species distribution data with Main conclusions: Current and future potential distribution of L. lucidum overlaps with biodiversity hotspots worldwide. Control of L. lucidum is a challenge once established. Thus, species distribution modelling helps to identify risk areas, guiding their early detection in current or future suitable areas. Our findings can be useful as a guide to develop region-specific invasion management strategies to prevent and/or control this species' spread.

K E Y W O R D S
biodiversity threat, biological invasion, ecological forecast, ensemble models, glossy privet, species distribution models environmental data, have been extensively used for different species (Guisan et al., 2017;Thuiller et al., 2019;Zhang et al., 2019).
Thus, here we aim to answer the following questions: (a) In which countries is L. lucidum present, and what is the current potential distribution based on climatic and edaphic variables? (b) Which areas of the World could, in future climatic conditions, exhibit suitable environment conditions? and (c) What is the current and future susceptibility of the biodiversity hotspots to be invaded?

| Data collection and curation
A total of 5,822 presence records were collated from a large number of herbarium specimen data from twenty-six Herbariums in six countries (Herbarium List.  Gavier-Pizarro et al., 2012;Hoyos et al. 2010;Presutti, 2011) and extensive personal observations made by L. Montti, N. Acosta-Aguirre, C. Fagúndez, M. Ayup, K. Garcia and S. Pacheco.
Records in native or novel areas were checked by screening available literature and by consulting with botanical experts. Records curation consisted of exhaustive revision and excluding records where information was considered unreliable, including potential misidentified specimens. We also excluded erroneous records (e.g. coordinates outside the studied area under which they had been listed, records located in the sea), and duplicates resulting from the merger of records from different sources. Furthermore, we excluded records with coordinates without decimals or with half-degree decimals (i.e. 0.5), georeferenced in the countries or states centroid, capitals or institutions. The native/invasive status in each country was classified according to literature and botanical expert consultation. After this process was completed, we ended up with 1,959 presence records.
Species' records are frequently over-represented in highways, urban areas, and protected areas, among others (Mccarthy et al., 2012;Meyer et al., 2016), and this has the potential to reduce a models' quality (Beck et al., 2014) if it is not accounted for. We chose to correct this records sampling bias by using environmental filters.
This method consists of creating a regular grid that divides the environmental space and then filters the records by the random selection of one occurrence within each grid-cell (Varela et al., 2014). An alternative is to use spatial filters (i.e. performed in the geographical space), but this has been demonstrated to be less efficient than the environmental one (Castellanos et al., 2019). We used an environmental filtering protocol adapted from Velazco et al. (2020). This consisted of performing a principal component analysis with environmental variables (see environmental variable section) and used the first three principal components to describe environmental space where occurrences will be filtered. Then, regular grids were created in this multivariate space with different cell sizes. The cell size is determined by the number of bins where each variable range is divided (Castellanos et al., 2019). There is a trade-off between the number of bins and the number of filtered records because as the number of bins decreases, the cell size of the grids increases, and the number of filtered records decreases (Castellanos et al., 2019).
In order to access the optimum number of bins, we tested five number of bins between 10 and 50. For each one, we calculated the spatial autocorrelation among filtered records based on Moran's I (Moran, 1948) and the number of filtered records. Then, we selected the set of bins with the lower quartile of the Moran's I and, of these, the one with the highest number of records .
The best number of bins was 50, which had a mean Moran's I of predictors of 0.447. Finally, we obtained 957 cleaned and unbiased records, which were used to construct species distribution models.

| Environmental variables
We chose to use climatic and edaphic variables to create SDMs because their combination has been demonstrated to increase model performance for plants (Bertrand et al., 2012;Thuiller, 2013;Velazco et al., 2017). We used 19 bioclimatic variables for current period (1979-2013) and two periods, 2050 (2041-2060) and 2080 (2061-2080), for future climate conditions available in the Chelsa v1.2 database with 30 arc-sec resolution (~1km at the equator; Karger et al., 2017). CHELSA future climate data use a delta change method by B-spline interpolation of anomalies (deltas) to correct data bias . For edaphic data, we used six physical soil properties provided by the SoilGrids with 30 arc-sec resolution (Table S2; Hengl et al., 2017). We selected these variables, assuming they are more stable through time than those related to chemical edaphic properties (like carbon stock or pH). Climatic and edaphic variables were upscaled by calculating their mean at 0.2 degrees resolution (~22 × 22 km).
Because knowing the geographical distribution of the species projected under future climate scenarios can help guide risk management strategies, we evaluated the potential invasion assuming two Representative Concentration Pathways, RCP 4.5 (moderate emission scenario) and a pessimistic emission scenario RCP 8.5 (IPCC, 2013;Meinshausen et al., 2011). The RCP 4.5 assumes the imposition of emissions mitigation policies, with moderate global emissions of greenhouse gases and land-uses, which stabilize radiative forcing at 4.5 W/m 2 (approximately 650 ppm CO 2 -equivalent) and global mean surface temperature exceeding 1°C by 2,100 . The RCP 8.5 represents one of a suite of scenarios that describe the highest greenhouse gas emissions combined with a high population and relatively slow income growth, with modest rates of technological change and energy intensity improvements . The RCP 8.5 delivers a global mean surface temperature increase of about 4.3°C by 2,100, relative to pre-industrial temperatures, a potentially extreme future situation likely resulting from society's failure to take low efforts to reduce greenhouse gas emissions (UNEP, 2017). We used multiple Atmosphere-Ocean General Circulation Models (AOGCMs) in order to account for variability driven by RCP 4.5 and 8.5 projections. We used the AOGCMs:

| Species distribution models
The choice of algorithms is the primary source of uncertainty for current and future projected distributions generated by SDMs Gaussian Process (GAU, Golding & Purse, 2016). For RF, the model was tuned automatically using the 'tuneRF' function of the random-Forest R package (Liaw & Wiener, 2002) based on the improvement of the out-of-bag error parameter; thus, the best model used a mtry parameter equal to three (i.e. number of variables randomly sampled as candidates at each split). RF was performed with default values of the step factor, 500 trees. SVM was performed with a radial basis kernel, with a constant cost value equal to one, and based on probabilities classes. We employed an inhomogeneous Poisson process procedure (Phillips et al., 2017b) to Maxent fitting. We used linear, quadratic, product and hinge features, default regularization values, 10,000 background points and clog-log output format. The Gaussian Process (GAU) models were constructed using a Laplace approximation (Golding, 2014). For RF, SVM and GAU, we randomly created pseudo-absences distributed 50 km distant to each presence, resulting in a number of pseudo-absences equal to the number of presences ( Barbet-Massin et al., 2012). We used Terrestrial Ecoregions of the World (Olson et al., 2001), where L. lucidum was recorded to constrain the area used to construct SDMs, that is the area where pseudo-absences and background points were located.
We used presence records from native and invaded regions together to create SDMs. Recently, it was reported that this species might be experiencing a climatic realized niche shift in the course of the invasion process (Dreyer et al., 2019). This indicates the non-equilibrium of species niche and the possibility that the fundamental niche may not be represented adequately only in the native region (Andrade et al., 2019). Ligustrum lucidum is easily transported by humans, and as we mainly focus on predicting the areas susceptible to its invasion, we estimated habitat suitability and interpreted results assuming no limitations to their future dispersal.
Hence, the use of native and novel areas pooled provides the best option (Beaumont et al., 2009;Broennimann & Guisan, 2008;Zhang et al., 2020). Because we are also interested in projecting this model to a worldwide scale, we evaluated model performance using spatial block cross-validation because it controls for the potential spatial autocorrelation between training and testing data and evaluates model transferability more appropriately than other partitioning methods (Roberts et al., 2017). To select the best grid size ( Figure S1), we generated 20 grids with resolutions ranging from 0.5 to 10 degrees and selected the one with (a) the lowest spatial autocorrelation, (b) the maximum environmental similarity and (c) the minimum difference of numbers' records between training and testing data . The optimum selected grid had a grid size of 9.5 degrees, a Moran's I of 0,168, a MESS of 14.965 and a standard deviation between the number of occurrences between partitions of 0.181. It is prudent to validate SDMs performance based on multiple evaluation metrics (Sofaer et al., 2019) because they can be variable regarding their threshold dependence (Liu et al., 2009) and sensitivity to prevalence (Leroy et al., 2018). Therefore, we used the metrics Area Under the Curve (AUC), True Skill Statistic (TSS), Jaccard, Sørensen, F-measure on presence-background data (Fpb), and the Boyce Index (Allouche et al., 2006;Leroy et al., 2018). We used a threshold that maximized Sørensen and Jaccard metrics to calculate threshold-dependent metrics (i.e. TSS, Jaccard, Sørensen and Fpb).
The final model was constructed performing a PCA with the suitability estimation of the four algorithms both for current and future conditions. We used this ensemble approach because it reduces uncertainty derived from individual models (Zhu & Peterson, 2017). Because we constrained the area used to construct models and we aimed to perform global and future projections, it is likely that our models will estimate suitability values for out-of-range environmental conditions used to fit them (model extrapolation). We evaluated model extrapolation for both current and future conditions based on the Mobility-Oriented Parity (MOP) approach (Owens et al., 2013).
This approach calculates the mean Euclidian distances between the environmental condition of a given point where a model will be projected and each point of the calibration area (Owens et al., 2013).
One characteristic of MOP is it allows to restrict the proportion of the nearest portion of the cloud of calibration area used for calculation (we used 10% of the cloud of calibration area). Current and future species' ranges were constrained with a very conservative threshold assuming as potential invasion only those cells with MOP values ≥0.9 ( Figure S4). We used the standard deviation of different algorithms and AOGCMs as future models' uncertainty.
Data processing, models and figures were performed with R software v3.6.1 (R Core Team, 2019); the R packages 'raster' (Hijmans, 2017) was used to handle spatial data and 'CoordinateCleaner' (Zizka et al., 2019) in conjunction with our own scripts, was used to perform records cleaning and sampling correction bias. We used the 'ENMTML' R package to create, evaluate and project SDMs (Andrade et al., 2020). We used the 'randomForest' (Paradis et al., 2004), 'kernlab' (Karatzoglou et al., 2004), 'GRaF' (Golding, 2014) and 'maxnet' (Phillips, 2017a) packages to fit RF, SVM, GAU and Maxent models, respectively. We used goode homolosine projection (EPSG 8,802) to estimate the surface extent of suitable areas with a grid-cell resolution of 22 × 22 km. All analyses and computed descriptive statistics were conducted using Spatial Analysis tools from ArcGis 10.1 and R v3.6.1 (R Core Team, 2019). To evaluate invasion risk and susceptibility to invasion in global priority biodiversity conservation areas, we focused our analyses using the Mittermeier et al. (2004) classification of biodiversity hotspots.

| Global map based on presence records
Historical evidence from herbarium catalogues revealed that earlier itemized species records outside its native range were documented since the middle of 1,800. The number of records per year increased rapidly after 2000. Ligustrum lucidum started to be methodically registered later in Oceania compared with Europe and America, but today this continent presents the highest number of records.
Africa was the continent where the species was later recorded and currently provides fewer records (Figure 1). Australia, New Zealand, United Kingdom, Turkey, Georgia or Pakistan), the species has been introduced but is not considered a problem or not information about its status was declared yet (Figure 1, Table 2).

| Current potential distribution
The final ensemble model had a Boyce and Fpb metric higher than individual models indicating good performance in predicting the suitable area for the species (Table S4). The potential current distribution (i.e. potentially invadable) of L. lucidum around the globe is 14,201,846 km 2 , an area slightly smaller than the size of South America (17 million km 2 ; Figure 2). In China, the species has a broad potential distribution, with more than 2,489,214 km 2 of suitable environmental conditions. However, the centre and south of the country appeared to be the most suitable area, as are neighbouring regions within Laos and Myanmar. In the rest of Asia, southern parts of Japan and Korea present regions of environmental suitability too, and, consistent with this L. lucidum has been declared as a problem for native ecosystems in these countries (Table 1) (Tables 1, 2 and Tables S5 and S6). Under RCP 4.5 scenario from the current period, it is expected a 22.5% of expansion during 2050 and a reduction of 4% in 2080 periods (17,408 and 13,626 km 2 , respectively) ( Figure S2 and S3). Under the worst-case emissions outcomes (RCP 8.5) by 2050s, there will be an overall reduction of 4% in the total suitable area compared to current conditions, and this reduction may increase to 25% by 2080s. The total suitable areas projected to be available are 13,631,660 and 10,631,660 km 2 for 2050 and 2080 periods, respectively (Figures 3 and 4).

TA B L E 1
Countries where Ligustrum lucidum is considered as an invasive species and its current and future potential distribution assuming a RCP 8.5  (Table S5 and S6, Figure S3).
In the other hotspots of biodiversity ( Figure 5) Notably, climate change may increase the potential distribution of suitable areas and thus increase the risk for invasion in the Tropical Andes and Eastern Afromontane, two of the most biodiverse and endangered regions of the world ( Figure 5). A similar pattern was observed under RCP 4.5 ( Figure S5).

TA B L E 2
List of countries with current and future potential distribution of Ligustrum lucidum >10,000 km 2 assuming a RCP 8.5

| D ISCUSS I ON
We modelled the current and future potential distribution of L. lucidum at a global scale using a comprehensive records database from native and novel ranges. We showed that the suitable environment presents an extensive geographical distribution, which includes regions where the species is not currently known to be present or invading. While we find differences in the total suitable area between both RCP scenarios (overall reduction under RPC 8.5 and an increment under RCP 4.5), we identify regions that are consistently projected to become newly suitable regardless of the scenario modelled. Notably, some of the regions found to be currently suitable but not yet occupied, and some of the regions projected to become newly suitable under climate change overlap with important hotspots of biodiversity.
Based on the records available, we found that L. lucidum is already globally widespread. However, we also highlighted that the efforts and attention to record its early introduction and their impacts have been heterogeneous through time and among countries. Today, information is still incomplete, including in countries with high suitability areas ( Table 2). As has been reported to occur for other invasive species (Shackleton et al., 2019), the perception of the problem, if it exists, can depend not only on ecological criteria but also on the social and economic circumstances present within a country and, in particular, as a function of the research resources available for environmental science within a country. Consistent with this, we observed that developed countries such as the United States, Australia or New Zealand, presented the most complete records of L. lucidum and in case of invasion detected today, they have a strong capacity to research and manage the species invasion (Early et al., 2016) with good national conservation plans in place. In contrast, in other regions, including across large parts of Africa and America that our modelling suggests are likely to be suitable for the species, there is a paucity of invasion. While this may reflect a lack of invasion by L. lucidum to date, it may also be due to a lack of recording and research.
Our model predictions for suitable environmental outside of the native range are largely consistent with the species' current global F I G U R E 2 Continuous (a) and binary (b) current potential distribution of Ligustrum lucidum based on an ensemble of four algorithms. Grid-cells for which potential distribution was determined to be model extrapolation were eliminated based on MOP analysis in the binary map.
Habitat suitability in panel range from 0 to 1 distribution records (Figure 1). This was expected given that we constructed our model using presence data from native and non-native ranges. However, importantly, the predicted potential distribution is wider than the present distribution of the species. This may suggest that the invasion by L. lucidum has not yet achieved its full expansion range. Notably, the model indicates areas, such as the Andes and the Himalayas slopes, which are at risk of a potential invasion. In contrast, the model also identifies some environmentally suitable regions where we know that L. lucidum has been introduced and have not yet reported any invasion (e.g. Italy, the United Kingdom and Guatemala). While it might be that a subtle environmental factor that we did not incorporate in the model accounts for this, it could also be F I G U R E 3 Future suitability pattern of Ligustrum lucidum for the periods 2050 (a) and 2080 (b) assuming a RCP 8.5. These models were based on the ensemble of four algorithms and five AOGCMs. Habitat suitability in panel range from 0 to 1 F I G U R E 4 Change in suitability pattern of Ligustrum lucidum for the 2050 (a) and 2080 periods (b) assuming a RCP 8.5. Colours depict different types of distribution changes regarding current conditions, orange: stable areas, red: potential distribution lost, blue: potential distribution gains. Grid-cells considered as model extrapolation were eliminated based on MOP analysis in future suitability projection that other abiotic or biotic factors present in these regions restrict the ability of the species to naturalize and spread.
Ligustrum lucidum tends to grow mainly in warm temperate macroclimate and has expanded its invasion range towards wet tropical areas (Dreyer et al., 2019). However, future alteration of climatic patterns could influence its distribution, as has been predicted to occur with other invasive plants (Hellmann et al., 2008). Interestingly, the invaded areas could potentially protect the species and its genetic diversity.
Outside of the species' native range, the model projects that some regions will decline in suitability, while others will increase; this can be important in identifying those regions most at risk from F I G U R E 5 Potential distribution for Ligustrum lucidum (km 2 ) among high biodiversity hotspots for current   Although environmental species distribution models play a crucial role in predicting species' potential distribution (Andrade et al., 2019), we recognize that our predictions could be limited due to multiple uncertainties related to invasive species. A particular concern in using SDMs to project the potential distributions of invasive species is the recognition that climate niche shifts can occur during the process of invasion (Early & Sax, 2014;Medley, 2010;Parravicini et al., 2015). Thus, the spatial extent projected to be suitable for an invasive species may be substantially underestimated. We considered this issue in how we constructed the SDMs. Ligustrum lucidum has a relatively long history as an invasive and has been introduced to multiple regions worldwide. By building our SDMs from presence data from native and non-native regions, we have projected much broader regions of suitable space for the species than would have used only data from the species' native range. It is interesting that for China and some parts of Asia, to which it is native, when we project suitability based upon the model constructed with global data, we identify substantial regions as being suitable that the species does not currently occur within. It is likely that it was previously present in some of these regions, but it no longer occurs due to habitat loss and exploitation. However, we suggest that in other parts of China its realized niche may be substantially smaller than its fundamental niche due to competition from or predation by native flora and fauna.
Outside of China, enemy release (Keane & Crawley, 2002) may be enabling the species to realize a higher proportion of its fundamental niche. The degree to which that occurs likely depends upon the local flora and fauna in the introduced region (Allen et al., 2016;Meijer et al., 2016;Montti et al., 2016). Hence, some of the differences we observed between projected climate suitability and documented invasion may relate to the extent to which interactions with the local native biota, and other factors not considered here, that can constrain the potential niche.
The use of SDMs can play an important role in assessing regions' vulnerability to particular invasive species, but they also have some limitations that emerging approaches can help to overcome.
Importantly, and as discussed already in the context of enemy release, our models were based on abiotic factors only, with no explicit incorporating of biotic factors that determine the potential range of species (Wisz et al., 2013). Ligustrum lucidum apparently has high plasticity in ecophysiological traits that allows it to be preadapted to different environmental conditions. One potentially interesting avenue for future work would be to incorporate L. lucidum within dynamic models that explicitly represent demographic and dispersal dynamics (e.g. RangeShifter; Bocedi et al., 2014)  to explore the interplay between pollination services and land-use (Synes et al., 2018), and there is exciting potential in taking a similar approach for invasive species.
In conclusion, we developed models to predict potential suitable areas for the establishment and spread of L. lucidum outside its native area. The main results showed that both current and future potential distribution areas overlap with important hotspots of biodiversity across the world. Threats to hotspots are similar and potentially even more intense than the threats facing biodiversity worldwide. It is important to account for current and potential future spread of invasive plant species into hotspots as they can have massive ecosystem effects, wholly changing hydrology, biogeochemical cycles and, ultimately, biodiversity (Mittermeier et al., 2004). Ligustrum lucidum introduction is lead by humans and could invade different hotspots reducing their biodiversity (Fernandez et al., 2020). Hence, these areas should be considered a priority to early detection and control strategies, both under current and future climate change scenarios. Thus, we recommend that the scientific community, policymakers, and stakeholders apply a precautionary principle and design early-warning schemes to detect the emerging invasion risk of L. lucidum. Additionally, the implementation of public campaigns about the potential invasive behaviours of this species and local scale strategies can help to prevent the establishment and further spread of L. lucidum. Furthermore, global and regional sharing of information can help to the maintenance of countries alert lists and assist efforts to limit the future invasion of this damaging species.

ACK N OWLED G EM ENTS
We

CO N FLI C T O F I NTE R E S T
Authors declare no competing financial or conflict interests.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13303

DATA AVA I L A B I L I T Y S TAT E M E N T
Authors have made available codes used for occurrence data cleaning, environmental data processing, and spatial distribution model-

B I OS K E TCH E S
The team includes multidisciplinary interests as ecology, conservation and evolutionary aspects of species biology. Their research integrates humans, biodiversity and environment and how these can affect the ecosystem processes simultaneously.
They use species distribution model and geographical information systems tools to response question about invasive species, landscape ecology and conservation.