In total, we identified 1,347 alien species in Russia from which we selected the TOP-100 most dangerous IAS (see Introduction). We studied this TOP-100 IAS in Russia using a factographic database (FDB) of alien species. We regard the database of alien species as a factographic database because it is used for the collection and storage of important “facts” about each invasive species, including year of introduction, pathway of introduction, impact mechanism, impact output, native range, donor territory, etc (Suppl. material 2). We grouped the 100 species according to their taxonomy into 16 different taxonomic ranks: Bacteria (two species), Chromista (three species), Fungi (three species), Alveolata (one species), Vascular plants (29 species), Ctenophora (one species), Nematoda (two species), Mollusca (12 species), Crustacea (12 species), Insecta (15 species), Ascideacea (one species), Actinopterygii (five species), Amphibia (one species), Reptilia (one species), Aves (two species) and Mammalia (ten species) (Suppl. material 3).

In addition, we divided the species into five ecological groups: “microorganisms” (bacteria, chromists, fungi, alveoli, nematodes, 11 species), plants (vascular plants, 29 species), aquatic organisms (ctenophores, mollusks, crustaceans, ascidia, ray-finned fish, 31 species), terrestrial ectotherms (insects, amphibians, reptiles, 17 species) and endotherms (birds and mammals, 12 species).

The FDB of the TOP-100 species was used to assess parameters of the invasion process, including pathways of introduction, description of the native range, and to catalog which IAS have multiple impacts on hydropower, agriculture, forestry, fisheries and hunting area and human health. We also used FDB to estimate the accumulation rate of the TOP-100 IAS in Russia over time. The main trends in IAS introductions were determined based on the first record of establishment for each species in Russia. We used these data to construct regression model describing the dynamics of the number of IAS introductions over time.

A vector database (VDB) was created in ArcGis Desktop 10.6.1 (ESRI 2017) using our original field data on SOR in combination with literature sources and museum collections, mainly from the Zoological Museum of MV Lomonosov Moscow State University, the museum of the Zoological Institute, Russian Academy of Sciences, DP Syreishchikov Herbarium of Moscow State University, Depository of Living Systems “Noah’s Ark” of Moscow State University, and the Herbarium of the Botanical Institute of the Russian Academy of Sciences. Species occurrence records of the TOP-100 Russian IAS in North and South America, New Zealand, Western Europe, Southeast, South and Central Asia were obtained from the Global Biodiversity Database (GBIF) (www.gbif.org), Centre for Agriculture and Bioscience International (CABI) (www.cabi.org), Ocean Biodiversity Information System (OBIS) (obis.org), International Union for Conservation of Nature (IUCN) (www.iucnredlist.org), AquaMap (www.aquamaps.org), and early detection and distribution mapping system (www.eddmaps.org). SOR were collected from scientific papers and/or international sources of open access. We distinguished three types of SOR. The first type contained SOR described by accurate geographic coordinates. For this data type we first removed duplicates, and then we applied accuracy control filters (i.e. SOR that had accuracy of location of more than 5 km were excluded). Several taxa included in the TOP-100 list have subspecies rank. In this case, the SOR of the subspecies not listed in the TOP-100 were excluded. The second type included SOR whose locations are depicted in the maps without indicating accurate coordinates. For this type of data, geographic coordinates were identified using basemaps of Russia by procedures of geo-registration and binding the available species localities to basemaps using at least 30 control points chosen in the ArcGis Desktop 10.6.1 environment. Basemaps were obtained from Natural Earth of public domain map datasets (https://www.naturalearthdata.com/downloads/10m-cultural-vectors/10m-admin-0-countries). For the third type of data, we used only SOR that allowed us to determine the accurate geographic coordinates using GoogleEarth (earth.google.com) with an accuracy of at least 5 km.

We obtained a final set of SOR data by combining all three types of records after excluding duplicate records for species locations. Each SOR contained data collected over a period of twenty years (for some species, such as the Apodemus agrarius, over a period of 80 years) during which the species geographical distribution have been studied including species identity, accurate geographic coordinates and year of the first record. In total, there were 169,679 SOR, of which 100,613 were from the native part of the IAS range and 69,066 were from the invasive range (Suppl. material 3).

Spatial bioclimatic variables (BIOCLIM) with the numbers from 1 to 19 (Bio1–Bio19) were taken from a global dataset WorldClim 2.1 (http://worldclim.org/version2) with a resolution of 2.5 arc minute (~ 5000 m) (Fick and Hijmans 2017) for the historical climate. We used layers on a global scale because the native habitats of the TOP-100 IAS in Russia are located all over the world, including North and South America, New Zealand, North Africa, and Eurasia (Suppl. material 3). BIOCLIM variables characterize annual trends, seasonality, and fluctuations of temperature and precipitation that commonly affect species distributions (Root et al. 2003; Bellard et al. 2013). In the case of freshwater species, many studies indicated strong correlations between spatial structures and climatic variables (Jocque et al. 2010), mainly temperature and water access. BIOCLIM variables were used to construct SDMs for various near-water and freshwater species (Drake and Bossenbroek 2004; McNyset 2005; Bellard et al. 2013; Banha et al. 2017).

In addition, we created raster layers of environmental variables in marine environments using the MARSPEC databases (Ocean climate layers for marine spatial ecology) (Sbrocco and Barber 2013). These layers are required for construction of SDMs of marine species. MARSPEC databases contain information related to the topographic complexity of the seabed, sea surface temperature and salinity. All raster layers of environmental variables BIOCLIM and MARSPEC datasets were prepared with resolution of 2.5 arc minute in *.asc and *.geotif formats.

The selection of variables for the modeling was carried out using a two-step procedure. At the first step, we created raster layers for the model-training regions for each environmental variable using the BIOCLIM or MARSPEC datasets. The model-training regions were described by the minimal convex polygons that included occurrence records reported for the native and invasive ranges (Rodda et al. 2011). The training areas were chosen based on the native and invasive regions of the species location including Russian regions (European part of Russia, Asian part of Russia – Ural, Siberia, Far East), regions of North America, Central and northern parts of South America, Western Europe, Central, Eastern, South Asia. At the second step, variables from the two sets of data, BIOCLIM and MARSPEC, were tested. We then selected variables for models using the ENMtools R-package (Warren et al. 2010). Those variables between which the pairwise Spearman correlation coefficient was greater than 0.75 in absolute value were excluded. Multicollinearity was assessed by the VIF (Variation Inflation Factor) using the Usdm R-package (Naimi et al. 2014). The advantage of the Usdm package for VIF assessment was described by Guisan et al. (2017). The environmental variable was regarded as multicollinear and excluded from the model if VIF > 5 (Hair et al. 1995; Guisan et al. 2017).

This two-step procedure led us to select six environmental variables from the BIOCLIM (Fick and Hijmans 2017) and MARSPEC (Sbrocco and Barber 2013) databases, which were used to construct SDMs for land (terrestrial and freshwater) and marine species, respectively. Environmental variables from the BIOCLIM set involved Bio1 (annual mean temperature), Bio5 (max temperature of warmest month), Bio11 (mean temperature of coldest quarter), Bio12 (annual precipitation), Bio13 (precipitation of wettest month), Bio14 (precipitation of driest month). Environmental variables selected from the MARSPEC dataset included Bathymetry (depth of the seafloor), Biogeo05 (distance to shore), Biogeo08 (mean annual sea surface salinity), Biogeo11 (annual range in sea surface salinity), Biogeo13 (mean annual sea surface temperature) and Biogeo16 (annual range in sea surface temperature).

There is no well-established methodology for accounting for SAC to build models based on only occurrence records (i.e. presence-only data) (Elith and Leathwick 2009; Guisan et al. 2017). We used an approach which involved random filtration of SOR to establish minimum spatial distances between the records (De Marco et al. 2008; Nuñez and Medley 2011; Marcer et. al. 2012; Václavík et al. 2012). Afterwards, we tested for SAC of model residuals. Many IAS have a large number of SOR. Therefore, they differ by spatial density depending on the number of samples collected in the native and invasive ranges. For this reason, we selected SOR using the procedure of random filtration of subsamples using the spThin package (Aiello-Lammens et al. 2015). This procedure is designed to (1) decrease the spatial bias of SOR located in different parts of the species range and (2) reduce the SAC of residuals of the model using approaches described by Dormann et al. (2007). We prepared two groups of datasets for each species. The first group of datasets included all available SOR without filtration. The second group consisted of 17 datasets, each of which included randomly selected SOR with different minimum distances. For the creation of these datasets we used minimum spatial distances in the range from 35 to 595 km at 35 km step. The MaxEnt models (stages three and four see below) constructed using these datasets were then tested for significance of autocorrelation of model pseudo-residuals (1 – probability of occurrence generated by model) by calculating Moran’s I at multiple distance classes via the Ape (Paradis and Schliep 2019) and LetsR (Vilela and Villalobos 2015) packages. The significance was determined using permutation tests (Vilela and Villalobos 2015). We further used only those models that had a P-value greater or equal to 0.05.

Although the MaxEnt default parameters for SDM are based on a large set of empirical data (Phillips and Dudik 2008), recent studies have shown that these models can be ineffective (Radosavljevic and Anderson 2014; Halvorsen et al. 2016). For this reason, we used three threshold-independent evaluation metrics: AUC_Test, AUC_Diff, and the Akaike size-adjusted information criterion (AICc) to select optimal MaxEnt parameters using the R-package ENMeval (Muscarella et al. 2014). AUC_Test is a metric that measures the discriminatory ability of a species distribution model using SOR that were not included in model construction. AUC_Diff is the difference between the AUC, calculated from the AUC_Train of the training set and AUC_Test. This metric (AUC_Diff) is a measure of the overfitting of the model. High AUC_Diff values characterize the degree of overfitting of the model (Warren and Seifert 2011). AICc – Akaike information criterion adjusted for small sample size reflects the degree of goodness-of-fit and complexity (Guisan et al. 2017). The model with the lowest AICc value is assumed to be the most appropriate model among the set of possible models. To estimate AUC metrics, we separated the calibration and evaluation SOR using the geographically structured partitioning scheme proposed by Radosavljevic and Anderson (2014) and implemented in the ENMeval R-package (Muscarella et al. 2014) (Suppl. material 4) . This SOR partitioning scheme leads to more realistic and less biased estimates of SDM performance than the more traditional partitioning scheme using random k-fold. The scheme of geographically separated SOR to determine the MaxEnt parameters is quite effective (Title and Bemmels 2018).

The ENMeval package created a series of MaxEnt models for each dataset using different regularization values and feature classes, compared them using the AICc criterion, and selected the most appropriate model. This package commonly selects a model that is less complex than the default model accepted by MaxEnt with an acceptable value of AUC_Diff metric (Halvorsen et al. 2016; Title and Bemmels 2018).

The final IAS distribution models (SDMs) adapted to the historical climatic conditions were built using package MaxEnt 3.4.1 (Phillips et al. 2006) and Biomod v.2.0 R-package (Thuiller et al. 2021). The procedure for the selection of optimal parameters for these models was described above. SDMs were created as a result of 10 MaxEnt runs to randomly select test and training samples. Seventy percent (70%) of the SOR in the MaxEnt runs were used as training samples, and 30% of the records were used for testing models. For calibration of the model, we used from 1,000 to 10,000 randomly generated pseudo-absence (PA) points based on the number of SOR. We used a combination of the ‘random’ and ‘sre’ strategies for the generation of the PA points using the Biomod v.2.0 R package. Number of PA points was generated (as recommended by Barbet-Massin et al. 2012) according to the number (N) of SOR (if N ≤ 1000 then 1000 points were selected, else 10,000 were selected).

We assessed the model performance using the Boyce (B_ind) index (Boyce et al. 2002; Hirzel et al. 2006) with the Ecospat R-package (Cola et al. 2017). Boyce index lacks several of the drawbacks associated with AUC index (Lobo et al. 2008). It requires only data on SOR and measures how much the predictive models differ from random distributions. The advantages of using this index have been shown in several studies (Hirzel et al. 2006; Petitpierre et al. 2012; Bellard et al. 2013; Petrosyan et al. 2019a, 2020b). We calculated B_ind for each of 10 model runs for each species, and then we averaged the run values to get the final estimates. The rank of each environmental variable in the SDM was estimated based on analysis of variable contributions and the jackknife method using MaxEnt. The highest rank was attributed to those environmental variables that had highest contribution in the model, i.e. had high values (> 5%) of permutation importance (PI) and/or high percent contribution (PC) (Phillips et al. 2006).

At this stage, we selected a model with B_ind > 0.7 and with insignificant SAC of residuals to compare it with the other models (see Selection of environmental variables and minimizing SAC section). If the model was accepted in terms of residuals, it was projected to the whole territory of Russia.

First, we binarized the original SDM maps for the analysis. We transformed the probabilistic maps obtained with help of MaxEnt into binary suitable/non-suitable maps using the threshold maximizing the True Skill Statistics (TSS) (Guisan et al. 2017). Afterwards, we used the vector layer of geographic coordinates of the administrative regions of Russia to find the number of environmentally suitable regions for potential species invasion, using the “Extract Value” function in ArcGis Desktop 10.6.1.

To obtain the actual range of invasion, we converted vector polygons of the Russian regions (including coastal areas of the seas) into raster giving unique numbers to each polygon. From this, we then determined the regions which contained at least one occurrence record. Raster binary maps of the species occurrences were created for each species in the regions of Russia. Further, using the vector layer of geographic coordinates of the administrative regions of Russia, we found the number of regions where each species has been already naturalized using the “Extract Value” function in ArcGis Desktop 10.6.1.

Hereafter we used three categories to interpret the type of the distribution of species based on the SOR – local, regional, and wide. Species with local distributions occurred relatively close to the sites of introduction and were established in less than 33% of the regions of Russia. Species occupying more than 33% but less than 66% of the regions of Russia were considered as regionally distributed. Species with wide distributions had almost continuous distributions colonizing more than 66% of the Russian regions from the western to eastern borders of Russia.

To find the actual number of IAS, we summed the created raster binary maps showing occurrence sites of species. Afterwards, we obtained the final raster map of the species richness in different Russian regions using the R-package raster. We then created a vector map of species richness in Russian regions for hot/cold spots analysis using the final raster map of IAS richness and vector layer of geographic coordinates of the centers of Russian administrative regions.

To identify zones with high impacts of the most dangerous IAS on Russian ecosystems, we used an analysis of hot and cold spots and a procedure of constructing kernel density in the ArcGis Desktop 10.6.1 environment. Hot, cold and neutral spots were identified based on the Getis-Ord Gi * statistic (Andy 2005) using Z-scores and P-values. As a result, we identified three types of zones distinguished by IAS richness that we regard as a proxy of IAS impacts. The first type of zone aggregates hot spots with the highest richness of IAS, which, consequently, have a great impact on ecosystems. The second zone type aggregates cold spots indicating regions with few invasive species that have reduced impact. Neutral zone type does not belong to the above two types and are characterized by intermediate richness and intermediate impact on ecosystems. Finally, we constructed the original map of hot/cold spots with Russian territory zoning according to assessments of IAS impact on environmentally suitable ecosystems.

We constructed SDMs using R (R-version 3.6.2 2019) and R-packages Ape (Paradis and Schliep 2019), Biomod2 (Thuiller et al. 2021), Dismo (Hijmans et al. 2017), Raster (Hijmans et al. 2020), Ecospat (Cola et al. 2017), ENMeval (Muscarella et al. 2014), ENMtools (Warren et al. 2010), SpThin (Aiello-Lammens et al. 2015), LetsR (Vilela and Villalobos 2015) and Usdm (Naimi et al. 2014). In addition, we applied R-scripts presented in Hirzel et al. (2006) to assess the suitability of models using RSTUDIO v. 1.4.1106 software (RStudio 2020). The analysis of hot/cold spots, construction of the kernel density of hot/cold spots and visualization of the SDMs were conducted in the ArcGis Desktop 10.6.1 environment (ESRI 2017).

The TOP-100 IAS of Russia were represented by 16 taxonomic groups (Suppl. material 3), of which 78 species were found in five groups: vascular plants – 29, insects – 15, molluscs – 12, crustaceans – 12 and mammals – 10 species. The other 11 taxonomic groups include only 1–5 species each (Fig. 1). We identified 169,709 SOR for the TOP-100 IAS in their native and invasive ranges, which included 1,117 for bacteria, 2,448 for chromists, 2,350 for fungi, 87,497 for vascular plants, 1,371 for alveolates, 760 for ctenophores, 407 for nematodes, 12,855 for molluscs, 11,995 for crustaceans, 3,780 for insects, 419 for ascidians, 14,746 for ray-finned fish, 5,197 for amphibians, 1,762 for reptiles, 2,279 for birds and 20,726 for mammals (Fig. 1). The number of SOR for individual species in their invasive and native ranges is presented in Suppl. material 3.

Figure 1.

Species occurrence records (SOR) in the native and invasive ranges of the TOP-100 IAS in Russia, which were used to assess actual and potential IAS invasive ranges. The number of species in each of 16 taxonomic groups is indicated in the grey bars; NR and IR are the numbers of SOR in the native and invasive ranges, respectively.

The number of SOR varies widely between ecological groups. The highest number of SOR (58,921) was recorded in plants, which consists of 29 species. Most of the plant species have colonized natural, disturbed, and/or urbanized terrestrial and aquatic ecosystems in Russia (Vinogradova et al. 2009, 2018; Vinogradova and Kupriyanov 2016). American maple (Acer negundo) has the highest number of SOR (8,849). All occurrence records for this species of maple are located in the area from the western (including Kaliningrad Region) to the eastern borders of Russia.

A total of 3,610 SOR were reported from 17 species of terrestrial ectotherms, 3,231 SOR from 31 species of aquatic organisms, and 2,413 SOR from 11 species of terrestrial endotherms. Within these ecological groups, the highest number of occurrence records (2,691) was reported for the marsh frog (Pelophylax ridibundus), which are located in the southern Ural, Siberia and Kamchatka (Kuzmin 2012). The lowest number of occurrence records (921) was found for 11 species of “microorganisms”. Such a low number of SOR for “microorganisms” was mainly attributed to the hard detection and species identification relative to representatives of the other groups. The highest number of records (208) in this group was reported for wetwood (Pectobacterium carotovorum) which were registered in all the forest areas on Russian territory except for the cold regions of Siberia (Cherpakov 2017).

The greatest number of invasive species originated from America (45 species), which included North America (31 species), Central America including Mexico (3 species) and South America (2 species), the western coast of these continents with adjacent waters of the Atlantic Ocean (8 species) and the eastern coast and the Pacific Ocean (1 species) (Suppl. material 3). One species (Potamopyrgus antipodarum) originated from New Zealand. Cylindrospermopsis raciborskii originated from either African or Australian. We were not able to establish the origins of 10 species. The native ranges of the rest of the TOP-100 IAS (43 species) were Eurasia and its surrounding waters. IAS from East Asia contributed the second most IAS to Russia (18 species). Native ranges of three species were located in Southeast Asia; two species originated from the north of the Pacific Ocean and adjacent land. Two terrestrial and one marine species invaded Russia from the south of Eurasia, two species invaded from Europe, two species came from Central Asia, and one species from the Caucasus. The Ponto-Caspian Basin was an important source of aquatic IAS (6 species) (Suppl. material 3). Among the aforementioned groups of Eurasian species, the range of 31 species at least partly includes the territory of Russia. Species living in East Eurasia, including the Russian Far East, as a rule, did not expand their range westwards but they invaded into Russian regions through Western Europe.

We found that 62% of the TOP-100 IAS were introduced into Russia unintentionally (accidentally), a third (33%) were intentionally introduced, and 5% invaded mainly by self-dispersal. The list of the TOP-100 is mainly represented by IAS carried to Russia with ballast water (22 species), cultivated plants (16 species), fouling of ships (11 species), and traffic flows (10 species).

The dynamics of the number of introduced IAS over time showed that there was a nonlinear relationship between the number of invasive species and year of introduction from 1600 to 2018 (F = 2138, P << 0.01, R² = 98%) (Fig. 2, Suppl. material 3). The number of IAS that invaded Russia started to grow sharply in the middle of the 20^th century. As such, the introduction of 52% of the species from the TOP-100 list occurred over a 76-year period from 1946 to 2018. Moreover, we found a significant difference (Z = 9.5, P << 0.01) in the rates at which new species were introduced (Rsy) between the two periods, namely, 1600–1945 (Rsy = 0.14 species/year) and 1946–2018 (Rsy = 0.65 species/year) (Fig. 2).

Figure 2.

Number of IAS introduced during 1600–2018 in Russia (panel A – red) and accumulated number of species (AS) (panel A, B – blue dots). In the panel B blue curve shows nonlinear trend AS = exp (-5.73 + 0.000002536* Years^2); red lines depict 95% confidence intervals.

We compiled a list of invasive species that have the greatest impact on biodiversity, various sectors of the economy, and human health (Fig. 3). Among the TOP-100 species, 36 species displace native species and 37 significantly alter the ecosystem (Fig. 3A). Seventeen species compete with native species and/or are able to displace them and five species are involved in hybridization processes. Thirty-seven species have great impacts on human health and eight species have considerable effects on hydropower systems. Some species are harmful for agriculture (29 species), forestry (20 species), fishery (13 species) and hunting grounds (8 species) (Fig. 3B).

Figure 3.

Impacts of the TOP-100 IAS on biodiversity, economic sectors, and human health in Russia. Panel A shows number of IAS altering ecosystems and displaced native species B shows impacts of IAS on economic sectors in Russia.

After random selection of occurrence points, the minimum distances between the SOR points ranged from 133 to 555 km, depending on the degree of SAC for each species. This procedure reduced the number of SOR by 88%, eliminated spatial sample bias and SAC of residuals (Suppl. material 5). Furthermore, it created similar distributions of SOR in the native and invasive ranges. Moran’s I correlograms for different distance classes did not show any SAC in the residuals of any of the models (Suppl. material 6). The reduced SOR data and six environmental variables from each dataset were used to construct SDMs. These models demonstrated high performance in terms of the B_ind index which varied from 0.76 to 0.99 with the mean 0.94 (Suppl. material 7). Based on SOR and SDMs, the studied species were divided into three groups according to their type of distribution, specifically, species with local, regional and wide distributions (Suppl. material 7). Suppl. material 7 includes short descriptions of IAS current distributions in Russia based on SOR, in particular, the number of regions in which species have already established or can potentially invade, the types of actual and predicted species distributions and model performance metrics (Boyce index – B_ind). Most of the IAS (56 species) currently have local distributions, slightly less (26 species) are distributed regionally, and even fewer (17 species) have wide distributions. MaxEnt models predicted a local distribution for 48 species, a regional distribution for 31 species, and a wide distribution for 20 species. Thus, SDMs predicted a decrease in the number of species (by 8) with a current local regional distribution and an increase in the number of species (by 5) with regional and wide distributions (3), respectively. The species with the widest distribution from each of 16 taxonomic groups (Suppl. material 7) are represented in Fig. 4.

The maps in Fig. 4 show that among 16 species, MaxEnt predicted local distributions for four species (Fig. 4E, F, K, N), regional distributions for seven species (Fig. 4B, C, H, I, J, M, O), and wide distributions for five species (Fig. 4A, D, G, L, P). Suppl. material 7 based on SOR data shows that among 11 less numerous species taxa, 16 IAS currently have local distributions, two species have regional distributions, and four species have widespread distributions. However, species distribution patterns are different in taxa that have a large number of IAS. In particular, a large proportion of invertebrate species have local distributions (90%), while a large proportion of vertebrate species have wide distributions (56%). For plants, the majority (66%) of species have regional distributions. A total of 78 (± 16)% of the potentially suitable regions for invasion in Russia have already been occupied (Poc) by IAS as predicted by SDMs. For species with local, regional and widespread distributions, the Poc values showed that 69 (± 15)%, 87 (± 6)% and 93 (± 4)% of potentially suitable regions had been naturalized, respectively. Thus, the largest portion of potential expansion area into environmentally suitable regions is predicted for species with local distributions (31%) and the smallest percentage of potentially expansion area is expected for species with widespread distributions (7%). For species with regional distributions, the portion of species expansion area into suitable regions is intermediate (13%).

Figure 4.

Predicted potential distributions of IAS including one species from the 16 taxonomic groups using SDMs. MaxEnt models used for prediction were optimized in terms of feature class and regularization according to the AICc metric. These models were constructed using selected environmental variables from the BIOCLIM or MARSPEC datasets, where species in the panels A, D, G, L, P have wide distribution, species in the panels B, C, H, I, J, M, O have regional distribution and species in the panels E, F, K, N have local distribution. Marine species and sea areas are shown in E, F, K.

Species occurrence records identified a positive relationship between the time since introduction and the type of distribution (either local, regional, or widespread), which was assessed by Spearman’s rank coefficient (Src = 0.58, P << 0.01). Species with local (56 species), regional (26 species), and widespread distributions (17 species) were introduced on average 55 (± 5), 126 (± 8), and 190 (± 12) years ago, respectively. This average residence time for species with a widespread distribution is likely an underestimate because there are IAS that were introduced before the 16^th century (Suppl. material 3).

SDMs showed that the environmentally suitable habitats for IAS with local distributions are mainly located in the European part of Russia. Species with regional distributions are found in the European part of Russia (9 species) and simultaneously in the European and Asian parts of Russia (22 species). The environmentally suitable habitats for the IAS with widespread distribution are located in the European part, as well as in the Asian part of Russia (Suppl. material 7).

We determined the IAS richness in each administrative division of Russian territory and adjacent seas, by overlaying the SOR on the polygonal raster map (Fig. 5, Suppl. material 8). The map shows that hot spots and zones with high concentration of IAS are located in the European part of Russia (Fig. 5). In the Asian part of Russia, only cold and neutral spots were found (Fig. 5). The maximum number of species was identified in hot spots A1 (64 species, Krasnodar territory), A2 (60 species, eastern coast of the Black Sea), A3 (55 species, Leningrad region), A4 (52 species, Rostov Region) and A5 (51 species, Moscow).

Figure 5.

Hot/cold spots and zones of IAS impact on ecosystems. Cold spots with -3, -2, -1 bins reflect statistical significance with confidence levels of 99, 95 and 90% respectively, spots with 0 bins have no statistical significance. Hot spots with 3 bins reflect statistical significance with confidence levels of 99%. Symbol A denotes hot spots, symbol Z denotes zones with a high concentration of hot spots based on kernel density analysis procedure. The pink line indicates the border between the European and Asian parts of Russia.

In the European part of Russia, we distinguished five zones with high concentrations of hot spots (Z1 – Z5). Zones Z1 and Z2 are located in the central part of European Russia and include 23 and 7 hot spots, respectively. Zone Z3 lies in Central Ciscaucasia and includes 7 hot spots. Two smaller zones (3 hot spots in each zone) are located northwest of the European part of Russia including the adjacent part of the Baltic Sea (Z4) and in the Northwestern Caucasus including the adjacent part of the Black Sea (Z5). Cold spots are located in the north-east of the European Russia and Asian Russia (Fig. 5, Suppl. material 8). Statistical significance levels of hot, neutral and cold spots of IAS included in the TOP-100 list are confirmed by GiZScore statistics.

The invasion process in Russia has a long history. Nevertheless, a non-linear upward trend in the dynamics of the total number of invaded species has been observed only since the late 1940s showing three peaks during 1946–1962, 1970–1981 and 2000–2014. Since 1946, the number of recorded IAS has doubled, and further increases are expected. A similar rapid growth in the number of invasive species has been documented in other countries since the second half of the 20^th century, including Croatia (Nikolić et al. 2013), East Africa (Witt et al. 2018) and Romania (Sirbu et al. 2022). The above studies show that peaks in the numbers of newly introduced species occurred in Romania (Sirbu et al. 2022) and Croatia (Nikolić et al. 2013) in the 1950s, while in East Africa (Witt et al. 2018) and Russia, the maximum numbers of introduced species declined during 1940s–1960s. Common factors that facilitated the increase in introductions in the second half of the 20^th century in Russia included increases in traffic flows (Kalabekov 2020) and interstate communications. Among the 14 species (represented in the TOP-100 list) that invaded Russia from 1946–1962, seven species were introduced with ballast water, fouling of ship hulls or accidentally with other aquatic organisms; four species were accidentally introduced via land routes (transport and shoots of plant culture); two species were intentionally introduced, and one species – Castor canadensis – invaded Russia independently from the territory of Finland.

Most of the TOP-100 IAS in Russia originated from North and Central America (45%), and the Asia-Pacific region (32%). A total of 62% of the IAS were unintentionally introduced, 33% were intentionally introduced and 5% were self-dispersed. The ratio of the number of unintentionally: intentionally introduced IAS is 1.82, which is higher than this ratio for invasive plant species in Europe (0.59) (Lambdon et al. 2008). Such a low ratio in Europe is related to the fact that the main pathways of introduction of invasive alien plant species (52% of the total number of invasive plant species in Europe) are dispersing through gardening and the use of alien plants for ornamental purposes. The horticultural industry, in particular, ornamental horticulture, is regarded as an important pathway for introduction and dispersal of alien species (Drew et al. 2010). In the TOP-100 list of IAS in Russia, the number of invasive species introduced with cultivated plants accounts for 16%. The TOP-100 list also includes a large portion of species introduced with ballast water, transport flows, ship fouling and agricultural products (58 species) (Petrosyan et al. 2017; Dgebuadze et al. 2018).

The identification of regions with the greatest IAS impacts on ecosystems is important for controlling IAS and forecasting potential regions of ecological disaster. We focused on administrative divisions of the Russian territory because the most important functions of nature management and environmental protection are implemented mainly by regional or republican services. Dangerous IAS occur in all regions of the Russian Federation. The greatest taxonomic diversity of IAS was found in the central part of European Russia, where the highest concentration of hot spots is observed (Z1). That is related to better civil development of the region and, consequently, greater anthropogenic transformation of this territory (Zubarevich 2010). The second most important zone of high IAS concentration is the Middle Volga region (Z2). After the construction of hydraulic structures linking the basins of the Baltic, Black and Caspian Seas into a single system, corridors of aquatic species spreading between the north (Baltic Sea) and the south converge in the middle reaches of the Volga (Slynko et al. 2002) to increase the number of IAS in this region. In our opinion, this invasion corridor contributed to form two hot spots in the Baltic (Z4) and Black Seas (Z5) regions. The increase in the concentration of hot spots in the south of European Russia (Z3 and Z5) is associated additionally with more favorable natural conditions for IAS, mainly with warmer climate. The identification of hot spots zones based on SOR would allow decision-makers to purposefully minimize threats from IAS, constrain their spread and/or eradicate IAS from the regional pool.

We followed best practices in model development (see in Fitzpatrick et al. 2006; Broennimann et al. 2007; Dormann et al. 2007; Kühn 2007; Beaumont et al. 2009; Guisan et al. 2014; Muscarella et al. 2014; Radosavljevic and Anderson 2014; Guisan et al. 2017; Petitpierre et al. 2017; Liu et al. 2020; Pili et al. 2020) to create SDMs for terrestrial, freshwater, and marine IAS. As a result, we were able to predict IAS distribution with high accuracy and without SAC in the residuals. Therefore, we strongly suggest using SDM and associated maps to predict expansion of IAS ranges into suitable regions.

Further species dispersal to environmentally suitable regions in accordance with predictions of MaxEnt models, is highly probable because measures to prevent self-dispersal and/or to restrict abundance were applied only for 19 species from the TOP-100 IAS in Russia (Dgebuadze et al. 2018). Control methods have been developed for 62 species including mechanical, chemical or biological approaches. Yet, there are no control methods to constrain population growth for 38 species either in Russia or in the world (Clout and Williams 2009; DAISIE 2009; Isaev 2015; Dgebuadze et al. 2018).