Climatic and anthropogenic factors shape the Asian range expansion of the invasive slug Arion vulgaris Authors: Ivan O Nekhaev, Anel A Ishayeva, Amina Omarova, Ersultan Zhuniszhanov Published: 2026 ABSTRACT AbstractThe Spanish slug Arion vulgaris is one of Europe’s most problematic terrestrial invaders yet, its distribution beyond the continent has remained poorly resolved. Outside Europe, the species has only sporadically been reported from North America, the Caucasus and a few isolated sites in the Southern Hemisphere, and its presence in most of Asia has been considered unlikely by previous climate‐only models. Here we combine targeted field surveys, integrative taxonomy and ensemble species distribution modelling to reassess the extra-European range of Arion vulgaris and to understand how climatic and anthropogenic factors jointly shape its invasion potential in Asia. We provide the first reliably documented record of Arion vulgaris in Central Asia, based on concordant morphological and COI-based genetic evidence from a population in the northern Tien Shan. We then construct a global niche model for the Arion vulgaris–Arion rufus–Arion ater complex using climatic predictors and land-cover fractions representing intensive agriculture, mosaic agro-natural landscapes and urban areas, and analyse predictor importance and partial dependence across a grid of Asian subregions with a European control area. Climatic variables dominate the overall model, constraining the broad potential range of the species, but in Central Asia, including the Western, Northern and Central Tien Shan, urban land cover emerges as a key predictor, allowing high predicted suitability in settlement mosaics embedded within otherwise climatically marginal mountain and continental environments. Compared with climate-only projections, our model substantially expands the area classified as suitable in Asia. These results imply an ongoing and largely undetected eastward expansion of the Arion vulgaris–Arion rufus–Arion ater group and highlight the need to interpret niche models in the context of land-use patterns when assessing invasion risks and potential impacts on endemic montane faunas.Keywords: Alien species, Arionidae, COI, Tien Shan, Niche modelingSubject terms: Climate sciences, Ecology, Ecology Introduction Biological invasions resemble natural colonisation in that both processes involve range expansion and integration into existing communities 1. Yet, invasions are typically accelerated and structured by anthropogenic disturbance, with establishment frequently occurring through human-altered habitat mosaics 2. For this reason, invasive species offer an opportunity to examine how climatic constraints and landscape structure jointly shape species distributions and ecosystem assembly. The Spanish slug Arion vulgaris Moquin-Tandon, 1855 is one of the most successful invasive invertebrates. It is included in the list of the one hundred most dangerous invasive species worldwide and is considered among the most successful invaders in Europe 3. The species is omnivorous, but primarily feeds on green parts and fruits of cultivated and wild plants 4. In areas where it occurs, the slug causes substantial damage to agriculture and can remove the natural herbaceous cover 5. In many newly colonised ecosystems, Arion vulgaris becomes a dominant species, displacing native terrestrial molluscs through both direct competition and hybridisation 6,7. Natural enemies of the species are scarce and include beetles of several species belonging to Carabidae and Staphylinidae 8, which prey on eggs and juveniles and can reduce slug abundance in agricultural habitats. Hedgehogs, amphibians, reptiles and some bird species may also feed on Arion vulgaris 9. Until recently, the expansion of Arion vulgaris was largely confined to Europe. The native range of the species remains debated. It was traditionally considered to originate from the Iberian Peninsula; however, recent molecular and phylogeographic studies suggest that it may be native to Central or Central–Western Europe 10,11. At the same time, alternative hypotheses based on population genetic data have also been proposed 12, indicating that the exact region of origin is still unresolved. At present, Arion vulgaris has been recorded in nearly all countries of Central and Northern Europe, in large parts of Eastern Europe 12–14. Outside Europe, only a few occurrences have been reported, in Canada, Mexico 13,15. The species was first documented in the Caucasus in 2019 in North Ossetia 16, and in 2022 it was recorded in Armenia 17. The latter represents the first reliable confirmation of Arion vulgaris in Asia. Until recently, no reliable records of the Spanish slug had been reported in Asia outside the Caucasus region. However, since 2013, large slugs visually resembling Arion vulgaris, and apparently misidentified based on external appearance as Arion rufus, have been repeatedly observed in the Botanical Garden of Almaty 18. These slugs most likely arrived with intentionally introduced plants 18. Since 2020, observations of large slugs similar in appearance to the Arion vulgaris group have also begun to appear on iNaturalist from Western Siberia and adjacent parts of Central Asia. Since 2023, slugs visually attributable to the group of large Arion spp. have also been recorded on Sakhalin, however, no precise identification has been carried out 19. Identification of Arion species cannot be performed reliably using external characters alone and remains challenging even with anatomical and genetic data due to interspecific hybridization within the group 7,20. Several morphological and genetic groups that are consensus-recognised as distinct species, including Arion vulgaris Moquin-Tandon, 1855, Arion rufus (Linnaeus, 1758) and Arion ater (Linnaeus, 1758), lack a clear reproductive barrier 7,20,21. As a result, they show high morphological variability and inconsistent identifications based on anatomy or individual genetic markers 7. Previous assessments of the potential distribution of this species group indicated that their occurrence is unlikely in most parts of Asia 13. These predictions relied exclusively on climatic variables 13. However, invasive populations of Arion vulgaris typically inhabit anthropogenically modified environments, where habitat conditions differ markedly from those in zonal ecosystems. In such environments, the slugs occur in gardens, vegetable plots, cultivated fields, household yards, parks and cemeteries 4,22. Through colonisation of human-modified habitats, Arion vulgaris can subsequently penetrate natural ecosystems 12,23. Field surveys conducted in 2025 confirmed the establishment of Arion vulgaris in the Tien Shan region of Central Asia, a finding not predicted by climatic niche models. We hypothesise that the presence of the species in this area is facilitated by the extensive anthropogenic transformation of local landscapes. In this study, we provide the first reliable identification of Arion vulgaris from the Central Asia and model the potential distribution of the species both globally and for Central Asia, including adjacent regions of the Caucasus, Siberia and Southeast Asia. Our modelling incorporates climatic factors as well as indicators of anthropogenic pressure. We further assess the likelihood of continued expansion of Arion vulgaris within Asia. Results Morphology The examined specimens measured 8–10 cm in body length when fully extended and 1–1.5 cm in width. The body colouration ranged from brown to reddish-brown, becoming darker towards the posterior part of the body. The mantle was oval with a granular surface, and the pneumostome was located in the anterior half of the mantle. The reproductive system generally corresponds to the morphology described for Arion vulgaris by Reise et al. 7. The genital atrium is clearly differentiated into a lower (distal) and upper (proximal) part (Fig. 1). The proximal atrium receives the openings of the epiphallus, oviduct, and bursa copulatrix duct. The epiphallus is relatively robust and curved. A distinct dilated portion of the oviduct is present between the atrium and the true oviduct (Fig. 1). Dissected reproductive system of Arion vulgaris (IZK-Mol-535). Abbreviations: bd – bursa duct, ep – epiphallus, la – lower atrium, do – dilated oviduct, ua – upper atrium. The COI sequence of this specimen is available in GenBank under accession number PX589502. Genetic analysis The two specimens collected in the Tien Shan region of Central Asia, for which COI fragments were amplified and sequenced, shared an identical haplotype that was also identical to a sequence originating from a cemetery in eastern Germany, as reported by Reise et al. 7 (Fig. 2). The clade containing the Tien Shan specimens also includes individuals previously assigned to Arion vulgaris in earlier studies 7,13. Maximum likelihood and Bayesian inference phylogenetic trees showing the phylogenetic position of the specimens collected in the Ile Alatau (red diamonds). The Maximum Likelihood (ML) and Bayesian Inference (BI) trees show differences in topology, mainly in the relative arrangement of clades containing sequences attributed to Arion vulgaris, Arion ater and Arion rufus. Low statistical support for several internal branches was likewise observed in previous phylogenetic studies of Arion based on multigene phylogenetic approaches or on COI alone 10,13,24. Modelling Comparison of models built with different buffer sizes (25, 50, 100, and 200 km) revealed only minor differences in predictive performance, as assessed by AUC, TSS, LOOIC and AIC (Supplementary Material). Metric values were largely similar across all buffer sizes, with small, inconsistent variations among them. Given the absence of a clear performance advantage, the 25 km buffer was selected for subsequent analyses as a conservative option that limits the background to conditions proximal to known occurrences. However, this main model showed a substantial improvement in evaluation metrics compared to models based solely on climatic predictors and to the variant using spatially shifted blocks (Supplementary Material). The block-shuffle models yielded lower AIC values, likely because AIC reflects model fit to the training data rather than predictive performance; the randomized predictors can still be fitted efficiently, while their ability to generalize is reduced, as evidenced by poorer AUC, TSS, and LOOIC values. Знaчeния AUC для выбpaннoй мoдeли cocтaвили (± standard deviation) 0.96 ± 0.001 для aлгopитмa LGM, 0.88 ± 0.003 для GMB, 0.97 ± 0.001 для MaxEnt, 1 ± 0 для RF и 0.98 ± 0 для вceгo aнcaмбля; знaчeния TSS: 0.82 ± 0.003 для GLM, 0.88 ± 0.003 для GMB, 0.84 ± 0.005 for MaxEnt, 0.98 ± 0.001 для RF и 0.86 ± 0 для вceгo aнcaмбля; AIC для GLM cocтaвил 7695 ± 118. World projections of the climate-only model, is provided in the Supplementary Material. In all modelling algorithms, climatic predictors dominated the explanatory power, contributing 96%, 90%, 85%, and 81% of the total variable importance in the Generalised Linear Model (GLM), Generalised Boosted Model (GBM), Maximum Entropy model (MaxEnt), and Random Forest (RF) models, respectively. Among non-climatic predictors, land-cover layers showed moderate effects (accounting for 4–19% of total importance), with urban and cropland fractions consistently ranking highest within this group. Complete variable-importance statistics are provided in supporting information on 10.5281/zenodo.17746140. Projections of the models produced at a resolution of 2.5 arcminutes for the global scale and 0.5 arc-minutes for Central Asia are shown in Fig. 3. In addition to the regions where the presence of Spanish slug populations has already been confirmed, our projection indicates high habitat suitability for members of the Arion vulgaris–Arion rufus–Arion ater group across a vast area of Asia, including the Himalayas, much China, the southeastern Pacific coast of the region, and several isolated foothill areas of the Western and Nothern Tian Shan. A high probability of invasion is also predicted for most of the Caucasus (Fig. 3), where Arion vulgaris has so far been recorded only sporadically 16,17. Model projections of the potential ecological niche of large Arion spp. at the global scale, for Asia, the Caucasus and the Tien Shan. Permutation importance (PI) indicated a strong influence of urban land cover on the predicted distribution of the Arion vulgaris–Arion rufus–Arion ater group in several Central Asian subregions, particularly in regions A2, A3 and A7 (Fig. 4). In some southeastern subregions, temperature-related predictors, especially Bio11 (mean temperature of the coldest quarter), showed a comparatively higher contribution. In the European part of the range, corresponding to the suggested native distribution of the Spanish slug 10,11 and used here as a control region, PI values for all predictors were uniformly low. Heatmap of permutation importance (PI) and graph of partial dependence (PD) values for the predictors used in the models and partial dependence plots for these predictors across the Asian subregions and the European control region. Region numbers correspond to those shown on the map. Partial dependence (PD) curves showed pronounced marginal responses for most climatic predictors in Europe and in one southeastern subregion of Asia (Fig. 4). In most other Asian subregions, the PD profiles remained nearly flat across predictor gradients, indicating a weak model response. The main exception was the urban-cover layer, which exhibited a clear and consistent positive effect in several Central Asian subregions, including areas of the Western, Central and Northern Tien Shan. Discussion Both morphological and molecular–phylogenetic evidence allow a confident assignment of the slugs collected in the Ile-Alatau of the northern Tien Shan to Arion vulgaris. This record represents the first reliable occurrence of Arion vulgaris in Asia outside the Caucasus and is located more than 2,500 km from the nearest previously confirmed locality 17. It also constitutes the easternmost confirmed occurrence of the species on the Eurasian continent. It should be noted that slugs likely belonging to Arion vulgaris have been reported even further east, on Sakhalin Island. However, in the absence of detailed morphological descriptions or genetic evidence, this record cannot be considered fully confirmed 19. In addition to the population analysed here, observations of visually similar large slugs have been reported on the iNaturalist platform and from several other sites within the Ile-Alatau area. We consider it highly likely that these observations also correspond to Arion vulgaris. We further cautiously suggest that this species is the most probable identity of similar slugs recorded from settlements across Central Asia and the southern part of Western Siberia. Overall, our ecological niche modelling indicates a strong dominance of climatic predictors in determining the potential distribution of large Arion spp.. However, across large parts of Central Asia, urban land cover exerts a substantial influence on model predictions. Surprisingly, the predictor representing partially modified landscapes, such as gardens, small-scale agriculture and low-intensity cropland, showed only a comparatively weak effect. Compared to the climate-only model (Supplementary Material) 13, the inclusion of anthropogenic land-cover predictors substantially expanded the area classified as environmentally suitable in Asia, particularly in regions with dense settlement mosaics. These differences were especially pronounced in Central Asia, where climate-based projections alone predicted almost no suitable habitat, especially in the Western, Central and Northern Tien Shan foothills where Arion vulgaris has been found. Also, in contrast to the previous modelling effort 13, our results indicate a high risk of further invasion of large Arion spp. in the Caucasus and more eastern regions, especially in China and parts of Southeast Asia. For the latter region, the high invasion risk is further supported by PD curves that are similar in shape to those observed in Europe, indicating a high level of environmental suitability across all examined predictors. The main limitation of our modelling approach, which may lead to an underestimation of the true distribution of the Arion vulgaris–Arion rufus–Arion ater group in Asia, lies in the use of occurrence data from iNaturalist. Such records are typically concentrated in human-visited locations, which can result in underestimating the influence of factors other than urbanisation, including agricultural land use. The current projection therefore relies, to some extent, on the assumption that European climatic conditions and land-use patterns represent the most favourable environment for the species. Additionally, the absence of natural enemies of large Arion spp. in Asia may facilitate a shift beyond the “European” ecological niche in the invaded region. A further limitation is that the model treats Arion vulgaris, Arion rufus and Arion ater as a single niche unit. Although Schoener’s D values indicate strong climatic overlap and do not support clear differences in environmental preferences among these taxa, combining them may slightly broaden the estimated niche envelope. This uncertainty affects only large-scale projections, not the identification of the Tien Shan population, which is independently confirmed as Arion vulgaris. On the other hand, the projection aligns well with the known physiological limits of the species. For adult individuals and egg masses, the experimentally demonstrated lower temperature threshold for survival is − 3°C 25. Egg masses also fail to survive at + 25°C 26. Eggs and juveniles are extremely sensitive to humidity and begin to desiccate even at relative humidity below 99.5% 27. These thresholds restrict the potential range of the slugs to areas with comparatively mild climates and render much of Central Asia unsuitable due to its sharply continental conditions, which fall outside these physiological ranges 28. Although minimum temperatures across much of the known distribution of the Arion vulgaris–Arion rufus–Arion ater group, including its native range, are considerably lower than − 3 °C, the lower survival limit reported for the species, microhabitat conditions near the ground may differ substantially from air temperature, which is typically colder than the soil 29. This allows organisms that are unable to tolerate freezing to survive even when air temperatures drop to low values. This assumption is further supported by verified records of the slug in northern Norway as far north as 69°N 14, as well as by its occurrence in the Caucasus 16,17 and the record from the Tien Shan described in the present study. Despite the limitations of the available occurrence data, we conclude that our projection provides a broadly accurate representation of the potential ecological niche of large Arion spp.. Nevertheless, the model could be further refined at the regional level as additional data become available. Our findings demonstrate that incorporating anthropogenic predictors can substantially improve invasion-risk forecasts. Most current studies modelling potential ecological niches within invasive ranges rely primarily on climatic variables. However, urban environments and agricultural landscapes, which differ markedly from natural habitats, may be particularly vulnerable to colonisation by non-native species and may serve as corridors facilitating their spread into natural ecosystems. Including such predictors in species distribution models is therefore likely to enhance risk assessments for large Arion spp. and other anthropophilic invasive taxa. Among mollusks occurring in Central Asia and adjacent regions of Siberia, several species have distributions influenced by human activity or anthropogenically modified environments. These include slugs of the genus Deroceras, the freshwater snails Physella acuta (Draparnaud, 1805) and Planorbella duryi (Wetherby, 1879), the Chinese pond mussel Sinanodonta woodiana (I. Lea, 1834), and several other species 30–34. The area where Arion vulgaris was confirmed in this study is subject to increasing anthropogenic pressure. This pressure results from the expanding use of mountain landscapes for recreational and agricultural purposes 35, the development of road infrastructure and a growing extent of built-up areas. According to our modelling results, continued anthropogenic transformation of the landscape may facilitate further expansion of the species within the Western, Northern and Central Tien Shan. At the same time, the intensifying aridisation and climate warming projected for the region 28,35,36 are likely to become important factors limiting the future spread of the Spanish slug. The presence of Arion vulgaris near Almaty, one of the largest cities in Central Asia and a major transportation hub, creates favourable conditions for the accelerated dispersal of the species to other suitable areas in Asia through passive transport by road and rail. The invasion of the Spanish slug into the foothills of the Tien Shan may pose a threat to natural ecosystems. Although confirmed record and model projections indicate that Arion vulgaris–Artion rufus–Arion ater group is largely confined to anthropogenically modified habitats, isolated populations may temporarily occur in adjacent natural ecosystems 23. These ecosystems host several tens of endangered species that are protected at national and international levels 37–40. Despite the generally low population density in the Tien Shan, built-up areas are distributed in a sparse mosaic pattern. They include small villages, recreational facilities and agricultural complexes, all of which may function as local hubs supporting populations of the Spanish slug and facilitating its further penetration into natural habitats. The Spanish slug is known to displace ecologically similar species, although verified evidence to date concerns only competition between Arion vulgaris and other members of the same genus 6,21. In the Tien Shan, the taxa most likely to experience direct competition from Arion vulgaris are the endemic slugs of the genus Turcomilax – Turcomilax turkestanus (Simroth, 1898), Turcomilax natalianus (Michaelis, 1892) and Turcomilax tzvetkovi Likharev et Wiktor, 1980, which are comparable in size to large arionids and occupy similar habitats 33,37. The known range of Turcomilax tzvetkovi is restricted to only two gorges in the Ile-Alatau, both subject to increasing anthropogenic development for recreational purposes and located close to the confirmed occurrences of Arion vulgaris. Importantly, competitive interactions between invasive and endemic slugs in mountain ecosystems remain poorly studied, and the presence of Arion vulgaris in the Tien Shan highlights the need for further research in this direction. However, the outcomes of such biotic interactions may be less straightforward than simple competitive exclusion. A series of experiments has shown that another limacid slug, Limax maximus Linnaeus, 1758, can negatively affect Arion vulgaris, consuming its egg masses and even biting adult individuals 41. This suggests that interspecific interactions among slugs may involve complex antagonistic relationships. Generally, the ecological consequences of this invasion for local ecosystems remain to be assessed. Similar potential threats to endemic faunas may arise in other mountain systems suitable for large Arion sp., notably in the Himalayas and the Caucasus. In summary, we provide the first reliable confirmation of Arion vulgaris in Central Asia, representing a major eastward range extension. Our modelling shows that while climate determines large-scale suitability, anthropogenic landscapes, especially urban areas, substantially increase invasion potential in otherwise marginal regions. The presence of Arion vulgaris near Almaty, combined with ongoing urbanisation and transport connectivity, creates favourable conditions for further spread. These findings highlight the importance of incorporating land-use predictors into invasion-risk assessments and underscore the need for monitoring in mountain systems hosting endemic faunas. Methods Sample collection and morphological analysis Specimens were collected on September, 1 and 9 2025 during two field trips to the Ile-Alatau State National Nature Natural Park (43.17786° N, 77.05473° E) at 1511 m a.s.l. (Fig. 5). The park is located in southeastern Kazakhstan near the city of Almaty and forms part of the northern Tien Shan. Two permissions for sample collecting were obtained prior to the study. The first document is an official permit issued by the administration of Ile-Alatau State National Nature Park, authorising the authors of this study to conduct joint fieldwork and collect specimens within the park. The second document is a general permit issued to all staff of the Institute of Zoology (Almaty) by the Committee of Forestry and Wildlife of Kazakhstan. This permit authorises scientific fieldwork and specimen collection in the regions encompassing the Tien Shan mountains. The Committee of Forestry and Wildlife is the national authority governing all state nature reserves and national parks in the country. Map of the confirmed occurrence of Arion vulgaris in the Ile-Alatau (red diamond) and the occurrence records used for ecological niche modelling of Arion vulgaris–Arion ater–Arion rufus group (green circles) after thinning at a 25 × 25 km grid resolution. Slugs were found near a paved road and residential buildings (Fig. 6). The study area is affected by anthropogenic transformation, including residential structures, a transportation route and intensive recreational use. Although potential shelters (construction debris, stones) were present, no individuals were found beneath them. Arion vulgaris was not recorded more than 3 m from the stream bed. In the nearby forested area, neither large native terrestrial molluscs nor Arion vulgaris were detected. General view of the habitat of Arion vulgaris (A, B) and live slugs (C, D) in the Ile-Alatau. Specimens were collected by hand, relaxed in a 5% MgCl2 solution and after that were fixed in 96% ethanol and tissues from the distal part of the foot were preserved for molecular analyses. Dissections of two specimens were carried out following the method of Likharev and Wiktor 42. Species identification was performed under a Leica M125 stereomicroscope based on the characters of the reproductive system described by Reise et al. 7. Specimens are deposited in the Institute of Zoology, Republic of Kazakhstan, under accession numbers IZK-Mol-535 (4 specimens) and IZK-Mol-540 (10 specimens). Molecular techniques DNA was extracted from foot tissues of two Arion vulgaris specimens using the DNA Blood & Tissue Kit (Qiagen, Netherlands). A fragment of the Cytochrome oxidase subunit I gene (COI) was amplified using primers LCO1490 and HC02198 (BioBeagle, Russia) and 4 × AllTaq Master Mix (Qiagen, Netherlands), following the protocol of Folmer et al. (1994). Sanger sequencing in both directions was performed by AllGene (Kazakhstan). All laboratory procedures were carried out at the Institute of Zoology (Almaty, Kazakhstan). Sanger reads were checked and manually assembled in Ugene ver. 52.1 43. The resulting sequences are available in GenBank under accession numbers PX589502 and PX589505. For comparison, we used COI sequences from GenBank previously published in other studies 7,13. Sequence processing was conducted in R ver. 4.5 for Apple Silicon processors following a previously established pipeline 31,44. The alignment generated with MAFFT under default settings was filtered using gBlocks with the least conservative parameters. The final alignment consisted of 484 nucleotides. The optimal nucleotide substitution model was determined in R using the phangorn 45 package and was identified as TPM3u + G + I. The Maximum Likelihood (ML) phylogenetic tree was inferred in R with 100,000 bootstrap replicates to assess branch support. Bayesian Inference (BI) was performed in RevBayes 46. The analysis ran for twenty million generations with sampling every 1000 generations in three independent runs, each comprising six chains. A burn-in of two million generations with parameter retuning every 100 generations preceded the main analysis. Convergence of the chains was assessed in R using the coda (https://github.com/CODA) package. All scripts used in this study are available on GitHub: https://github.com/anisus3/Phylogeny. Program generated trees are available on Zenodo: 10.5281/zenodo.17746140. Data for modeling To model the potential distribution of the species group, we used GBIF occurrence data for three species: Arion vulgaris (10.15468/dl.hywdbm – 29,562 occurences), Arion rufus (10.15468/dl.e724ww – 39,465 occurences) and Arion ater (10.15468/dl.65xfmj – 33,316 occurences). The modelling was performed for the species group rather than for individual taxa, as these species lack reliable external morphological differences and cannot be accurately distinguished in poorly verified open-access records. They also exhibit substantial ecological niche overlap, quantified using Schoener’s D metric 13. Based on our dataset, Schoener’s D values were 0.55 for Arion vulgaris × Arion rufus, 0.53 for Arion vulgaris × Arion ater and 0.49 for Arion ater × Arion rufus. In addition to GBIF data, we incorporated visually verified iNaturalist (https://www.inaturalist.org) observations of large slugs resembling Arion vulgaris from Asia (accessed 20 October 2025). This dataset comprises 45 records, many of which originate from geographically proximate localities. All four datasets were merged into a single occurrence table and spatially thinned to a minimum nearest-neighbour distance of 25 km in order to reduce spatial autocorrelation. Thinning was performed using a geodesic distance-based greedy algorithm, retaining only one occurrence within a 25 km radius neighbourhood while preserving the original geographic coordinates. The final dataset included 2757 occurrence records (Fig. 5). A relatively fine thinning distance was chosen to reduce spatial clustering while still capturing occurrences within urban and agricultural landscapes, which typically exhibit a mosaic spatial structure. We used two types of environmental predictors. First, we included climatic variables from WorldClim 47. Experimental data indicate that the survival of egg clutches and juveniles is primarily influenced by lower and upper temperature thresholds, as well as by humidity 25,27. Reproduction occurs in spring and autumn 48, when the lower temperature threshold and moisture availability may represent the main limiting factors. We selected Mean Temperature of the Coldest Quarter (Bio11) because it better reflects the potential for successful overwintering of juvenile individuals than absolute minima (e.g., Min Temperature of the Coldest Month, Bio6), given the buffering effect of microhabitat conditions 29. The remaining bioclimatic variables were chosen following pairwise comparisons using Spearman’s rank correlation coefficient. From each pair of variables with an absolute correlation coefficient exceeding 0.75, one variable was randomly retained. The final set of predictors included, in addition to Bio11, one additional temperature-related variable, Mean Diurnal Range (Bio2), and three precipitation-related variables: Annual Precipitation (Bio12), Precipitation of the Driest Month (Bio14), and Precipitation Seasonality (Bio15). Overall, the resulting set of predictors used for climatic niche modelling is well aligned with the known physiological constraints of the species. Second, we used data from the Climate Data Record (CDR) and Interim CDR Sentinel-3 Land Cover (v2.1.1) for 2022, which provide up-to-date information on global land-cover fractions 49. The selected layers were grouped into three categories: 1. Intensive agriculture: Cropland, rainfed (codes 10, 11, 12) and Cropland, irrigated or post-flooding (code 20) (cropland strict hereafter layer); 2. Mosaic agricultural landscapes mixed with natural vegetation (codes 30 and 40) (cropland mosaic layer); 3. Urban areas (code 190). For each of the three aggregated categories, we calculated the total proportion of all corresponding land-cover types at the selected spatial resolution. Preparation and processing of environmental layers were performed in R using the terra 50, geodata (https://github.com/rspatial/geodata) and corrplot (https://github.com/taiyun/corrplot/blob/master/R/corrplot-package.R) packages. Potential niche modeling The distribution of the Arion vulgaris species group was modelled at a spatial resolution of 2.5 arcminutes (~ 5 km) globally. The trained model was then projected onto Central Asia and adjacent regions, including parts of Siberia, the Caucasus and the Indo-Pacific sector, using a finer resolution of 0.5 arcminutes (~ 1 km). A separate Asia-only model was not developed due to the limited number of occurrence records. Species distribution models were implemented using the biomod2 (https://biomodhub.github.io/biomod2/) framework in R. Four algorithms representing complementary modelling paradigms were applied: Generalised Linear Models (GLM), Generalised Boosted Models (GBM), Random Forests (RF) and Maximum Entropy (Maxent) 13. To reduce model-specific uncertainty, predictions from individual algorithms were combined into an ensemble model using a weighted-mean approach, with weights proportional to each model’s predictive accuracy 51. This ensemble provides a consensus estimate of environmental suitability that integrates complementary information from all modelling techniques. A total of 12,000 pseudo-absences were generated following the recommendations of Barbet-Massin et al. 52, who suggested that large pseudo-absence sets (10,000 or more) improve model stability for regression- and tree-based algorithms. Pseudo-absence points were randomly sampled within the buffered accessible area. We fitted models using four alternative buffer sizes defining the accessible area (25, 50, 100, and 200 km) to assess the sensitivity of model performance to background extent. To evaluate the added predictive value of anthropogenic predictors, we additionally constructed a climate-only model using the buffer size that yielded the highest model performance. To assess whether the apparent effect of anthropogenic predictors could arise from their broad-scale spatial structure (and potential co-variation with climate), we applied a block shuffle procedure (i.e. a spatially constrained permutation). Each anthropogenic raster was reprojected to an equal-area grid, partitioned into 300 × 300 km blocks, and the blocks were randomly reassigned across the region while preserving within-block values. This procedure maintains local spatial autocorrelation while disrupting large-scale spatial structure. The procedure was repeated 100 times for each predictor to generate null expectations of predictor effects. This approach follows the general logic of spatially constrained permutations used to account for spatial autocorrelation 53,54, but is here adapted to generate null predictor layers for SDM evaluation. We preferred this approach because it is computationally less demanding than spatial block cross-validation while still allowing us to assess the influence of spatial structure. Model performance was evaluated using three complementary metrics: Area Under the Receiver Operating Characteristic Curve (AUC), True Skill Statistic (TSS), and the Leave-One-Out Information Criterion (LOOIC). AUC and TSS were calculated from validation data extracted from BIOMOD2 model outputs, whereas LOOIC was estimated as a proxy based on the summed log-likelihood of validation predictions under a Bernoulli error distribution. For LOOIC, lower values indicate better model fit, while higher values of AUC and TSS reflect improved discriminatory performance. In addition, the Akaike Information Criterion (AIC) was calculated for GLM models based on calibration data; however, as AIC reflects goodness-of-fit to the training data rather than predictive performance, it was used only as a supplementary measure. All scripts used for modelling are available on GitHub (https://github.com/anisus3/ecological_niche_modeling_v2), modeling projections, AUC, TSS and LOOIC values and environment significance are available on Zenodo (10.5281/zenodo.17746140). Assessment of predictor contributions in Asia To assess the contribution of individual predictors to local model predictions, we divided the Asian study area into a regular grid composed of three latitudinal and four longitudinal bands (12 subregions in total). As a control, we used a European region of comparable size located between 40° and 52° N and 6° and 32° E, corresponding to the core European range of Arion vulgaris, which includes the putative area of origin of the species according to some studies 10,11. All climatic variables and land-cover layers were represented at a uniform spatial resolution of 0.5 arcminutes across all subregions, including the European control area. For each of the thirteen subregions, we calculated permutation importance (PI), which quantifies the reduction in predictive accuracy after randomly permuting the values of a given predictor and therefore reflects its relative importance for the model. We additionally computed partial dependence (PD) curves to quantify the marginal effect of each predictor on model output. PD curves describe how the model’s mean predicted suitability responds to variation in a single predictor while the effects of all other predictors are averaged over the empirical data distribution. For each subregion, PD curves were generated using 100,000 randomly selected pixels. All PI and PD analyses were performed in R using the biomod2 package, and visualizations were produced with ggplot2 (https://ggplot2.tidyverse.org). Raw data for both PI and PD analyses are available on 10.5281/zenodo.17746140. Map construction and final image preparation Maps were prepared in QGIS ver. 3.40 (https://qgis.org) using the Natural Earth Map Kit (https://www.naturalearthdata.com) as a background layer. Final preparation of programmatically generated figures was performed in Sketch ver. 77 (https://www.sketch.com), including the addition of backgrounds, labels, annotations and font adjustments. Basic photo processing (cropping and background removal) was carried out in Pixelmator Pro ver. 3.6 (https://www.pixelmator.com/pro/). Acknowledgements The study was performed in the framework of project “Information system development for wild animals’ cadastre maintaining of Western Tien Shan for the conservation and sustainable use purposes” supported by the Ministry of Higher Education and Science of the Republic of Kazakhstan (No BR24993060). Author contributions Ivan Nekhaev designed the study, developed the main concept of the manuscript, performed the anatomical examination of the slugs, conducted the ecological modelling, prepared the figures for publication and participated in the fieldwork. Anel Ishayeva contributed to the field collections, manuscript preparation and literature search. Amina Omarova carried out the laboratory molecular work and reconstructed the phylogenetic trees. Ersultan Zhuniszhanov first discovered the studied population of *Arion vulgaris* during the field survey. Funding The study was performed in the framework of project “Information system development for wild animals cadastre maintaining of Western Tien Shan for the conservation and sustainable use purposes” supported by the Science Committee of the Ministry of Science and Higher Education of the Republic of Kazakhstan (No BR24993060). Data availability The ecological niche model projections, variable-contribution values, model evaluation metrics, raw software-generated images and the code used to produce them are available in Zenodo (10.5281/zenodo.17746140). The code used for both species distribution modelling and phylogenetic analyses is also available on GitHub (https://github.com/anisus3). The datasets used for modelling are accessible on GBIF ( *Arion vulgaris* – 10.15468/dl.hywdbm, *Arion rufus* – 10.15468/dl.e724ww, *Arion ater* – 10.15468/dl.65xfmj). The newly generated sequences are available in GenBank under accession numbers PX589502 and PX589505. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The ecological niche model projections, variable-contribution values, model evaluation metrics, raw software-generated images and the code used to produce them are available in Zenodo (10.5281/zenodo.17746140). The code used for both species distribution modelling and phylogenetic analyses is also available on GitHub (https://github.com/anisus3). The datasets used for modelling are accessible on GBIF ( *Arion vulgaris* – 10.15468/dl.hywdbm, *Arion rufus* – 10.15468/dl.e724ww, *Arion ater* – 10.15468/dl.65xfmj). The newly generated sequences are available in GenBank under accession numbers PX589502 and PX589505.