Invasion Patterns of the Coypu, Myocastor coypus, in Western Central Greece: New Records Reveal Expanding Range, Emerging Hotspots, and Habitat Preferences

Abstract

The coypu (Myocastor coypus), a semi-aquatic rodent native to South America, has established invasive populations across North America, Asia, and Europe. In Greece, since its initial recording in 1965, the species has been rapidly expanding, forming sizable populations in northern continental regions. However, the extent of its invasion and the environmental drivers shaping its distribution and spatial patterns in western–central Greece remain poorly understood. Here, we address this knowledge gap, aiming to identify and map new coypu records, investigate the relationship between coypu presence and habitat characteristics, and analyze its spatial distribution. Between 2020 and 2023, we conducted 50 field surveys across the study area, documenting direct and indirect evidence of coypu presence. We integrated kernel density estimation, Getis-Ord Gi*, and Anselin local Moran’s I to identify spatial distribution patterns and hotspots of the coypu. Additionally, we analyzed environmental factors including land cover type, total productivity, and geomorphological features to determine their influence on habitat selection. Our findings reveal significant spatial clustering of coypus, with 12 identified hotspots primarily located in protected areas, and highlight tree cover density and productivity variability as key predictors of coypu presence. The suitability of western–central Greece for the coypu appears to be driven by extensive wetlands and interconnected hydrological systems, with hotspots concentrated in lowland agricultural landscapes, providing essential data to guide targeted management strategies for mitigating the ecological risks posed by this invasive species.

1. Introduction

The intentional or unintentional introduction of species beyond their native range, driven by globalization and facilitated through intensified trade and travel, has become a major driver of global biodiversity loss, alongside land use changes, species exploitation, climate change, and pollution [1,2,3,4]. Invasive Alien Species (IAS) represent not just an isolated ecological concern, but a global challenge impacting ecosystems, biodiversity, and human activities worldwide [5,6,7,8,9]. The alarming surge in IAS introductions [10], as highlighted by a substantial 40% increase over the past 40 years, coupled with a more than two-thirds rise in the number of IAS per country over the last five decades, underscores the pervasive threat that transcends geographical boundaries and raises profound concerns for the sustainability of natural and human systems at both insular and continental scales [10,11]. The multifaceted impacts of IAS, across all biological scales [4], are deeply unsettling, as IAS act as ecosystem engineers, reshaping the structure and dynamics of ecosystems, disrupting ecological balances, and altering fundamental environmental processes [12]. Notable examples include invasive rodents such as black rats (Rattus rattus) [13], muskrats (Ondatra zibethicus) [14], and North American beavers (Castor canadensis) [15]. Black rats, which have invaded islands globally, prey on native bird eggs and consume seeds, altering plant communities [16]. Muskrats, introduced in Europe, degrade wetlands by destroying aquatic vegetation and destabilizing riverbanks [17], while North American beavers, when introduced beyond their native range, modify water flow patterns and displace native species [18]. Beyond these ecosystem-level disruptions, IAS pose a direct threat to crucial ecosystem services [19], while they can drive native species toward endangerment or extinction through various mechanisms, including predation, competition, hybridization, and pathogen introduction [20,21,22].

Several global and regional regulatory frameworks have been established to address the challenges posed by IAS. Within the European Union (EU), Regulation 1143/2014 [23] has been a pivotal legislative instrument [24], prioritizing preventative measures, early detection, and rapid response mechanisms to contain and eradicate established populations [25]. After all, in the EU, an estimated 12,115 naturalized alien species pose diverse ecological and economic challenges [26]. According to the DAISIE database (www.europe-aliens.org (accessed on 30 August 2024)), approximately 10–15% of these species qualify as invasive, imposing enormous annual financial costs [23,27]. While impacts extend across various taxonomic groups, vertebrates, represented by 40 species on the European list of high-impact IAS [3], garner particular attention due to their comparable representation to plants and invertebrates and their demonstrably disruptive effects on native ecosystems.

From this set of alien vertebrates established in Europe, nearly one-fifth belong to the mammalian class, highlighting the disproportionate ecological impact of this taxonomic group across Europe [23,28]. This dominance of mammals is concerning, given that they constitute the majority of the most detrimental invasive terrestrial vertebrates in Europe, highlighting their substantial ecological impact [29,30]. At the same time, when considering the relationship between IAS and the protected areas (PA) within the EU, particularly within the Natura 2000 network, this issue becomes even more complex. Elevated levels of invasion in the Natura 2000 network [31], a cornerstone of European conservation efforts [32], illustrate the vulnerability of PAs to IAS [33], necessitating focused studies to mitigate their impact.

Within this context, the coypu or nutria, Myocastor coypus (Molinia 1782), a semi-aquatic rodent native to South America, exemplifies the challenges posed by invasive mammalian species, ranking 17th among the worst invasive species in Europe and 6th among mammals [34]. Thriving in a variety of environments, from wetlands to urban areas, this prolific breeder exhibits remarkable adaptability [35,36]. Noteworthy is the coypu’s reliance on natural dispersal from established populations [37,38], as trade and farming pathways have significantly declined following regulatory measures [39]. However, small-scale, unofficial farming or private trade may persist in some regions, where hobbyists occasionally breed and translocate coypus, with instances of intentional or accidental release into the wild [40].

The coypu’s adaptability extends beyond habitat preferences, with a lack of dependence on any single species for survival. It readily exploits diverse food sources within its invaded range [39], complemented by non-seasonal breeding and large litters, along with the potential for multiple litters per year under favourable conditions, enabling rapid population growth [41,42]. This complex interplay of factors underscores the coypu’s significant impact on invaded ecosystems, earning the coypu its place on the list of the “100 of the World’s Worst Invasive Alien Species” [43]. Beyond Europe [36,44], it has established invasive populations in many regions of North America, Africa, and Asia [39,44,45,46] as a result of its introduction for the fur industry [44]. Within the EU, until 2022, the coypu had established permanent populations in 17 Member States including Greece.

Greece, a global biodiversity hotspot [47,48], harbors diverse ecosystems increasingly challenged by IAS [49]. Despite substantial progress being made in documenting IAS distribution as well as impacts across taxa [49,50,51,52,53,54,55,56,57], data on the distribution of certain mammalian IAS remain limited. In fact, while some invasive mammals, such as the American mink, are relatively well-documented [58], others, including the coypu, require further investigation to accurately document their current distribution in Greece.

Historical records indicate the species’ presence since at least 1965, with initial documentation at Lake Agra in northern Greece [49,59]. The species’ introduction appears to be linked to farming initiatives and deliberate releases, including a 1964–1965 project by the Public Power Corporation, which introduced 230 coypus to Lake Agra to manage aquatic vegetation [60,61]. Coypu farming was further supported through state subsidies in the 1960s, promoting small-scale breeding efforts [62]. By 1972, reports suggest that the population in Lake Agra had reached approximately 5000 individuals [61], despite earlier predictions in 1966 predicting numbers as high as 30,000 [61,63]. Over the decades, the coypu has shown evidence of range expansion, with populations documented in northern Greece [59] and on the island of Corfu [64].

However, our understanding of the coypu’s distribution remains fragmented, with existing reports, primarily from citizen science platforms, offering only glimpses of its presence, while verified data remain limited. The fragmented nature of such data also raises concerns about potential underrepresentation of certain habitat types, as citizen science often skews toward easily accessible or highly visible areas. Comprehensive field surveys, therefore, become indispensable for capturing the full scope of its distribution, particularly in under-sampled regions and habitats.

This paucity of comprehensive data underscores the urgent need for targeted research to delineate its precise range and assess the extent of its invasion. Although the available evidence suggests that the coypu likely expanded from its initial introduction sites, with the southernmost confirmed population in western–central Greece along the southern shores of the Amvrakikos Gulf, significant gaps persist, particularly in understanding the environmental factors shaping its range expansion and spatial distribution patterns.

To address this, we focused our study on western–central Greece, a region characterized by diverse hydrological networks and ecological conditions that could facilitate coypu colonization, aiming to: (a) identify and map new coypu records to determine the current extent of its invasion; (b) investigate the relationship between coypu records and habitat characteristics to identify factors influencing its distribution; and (c) analyze its spatial distribution patterns to delineate potential hotspot and coldspot areas, offering insights into regions where coypu presence may intensify ecological pressure, thereby providing essential information to guide targeted conservation strategies aimed at mitigating the impacts of coypu on vulnerable habitats and native biodiversity.

2. Materials and Methods

2.1. Study Area

Western–central Greece, lies between latitudes of 38°18′ N and 39°10′ N and between longitudes of 20°44′ E and 22°02′ E. It covers an expanse of 5450 km², constituting 4% of Greece’s total surface area. It features contrasting flat and fertile alluvial plains and a mountainous interior, prominently marked by the mountains of Panaitoliko, Akarnanika, Petalas, and Arakynthos. This region stands out for some of its distinctive geographical features, particularly several freshwater lakes, including Trichonida, which is Greece’s largest natural lake with a surface area of 95.8 km², as well as the wetland complex comprising the Messolongi–Aitoliko lagoons and the deltaic formations of rivers Acheloos and Evinos in the southwest of the study area. Formed by the convergence of the Acheloos and Evinos rivers, this area is recognised as a Ramsar Convention site. Of significance to coypus, further artificial lakes have been created along the course of major rivers. The ecological significance of the area is underscored by eight Special Areas of Conservation (SACs) and four Special Protection Areas (SPAs), encompassing lakes (Lysimachia, Trichonida, Amvrakia, Ozeros), mountains (Panaitoliko, Varasova, Arakynthos, Akarnanian), and the Acheloos–Evinos–Messolongi complex (Figure 1). The climate of the area is wet Mediterranean with hot and dry summers and cool and wet winters. Mean monthly temperatures range from 8.3 °C in January to 27.4 °C in July, with an average annual rainfall of 919 mm in Agrinio [65].

2.2. Field Survey

From 2020 to 2023, we conducted a total of 50 field surveys to investigate coypu presence and distribution across the study area using a randomized grid-based approach. The entire area was divided into 280 grid cells (5 × 5 km) using ArcGIS 10.7 (ESRI Inc., Redlands, CA, USA), with 166 cells measuring 25 km² and 114 smaller cells adjusted to fit the geographical boundaries of the region (Figure 1). All grid cells, regardless of size, were included in the survey to ensure a comprehensive spatial coverage, and each cell was visited once in a randomized sequence to minimize biases associated with habitat type and accessibility. Equal survey effort was applied across all grid cells to reduce over-representation of high-probability habitats and adequately sample less suitable areas that could support transient or dispersing individuals.

Surveys were conducted during the late boreal spring (May–June) and the early boreal autumn (September–October), targeting periods that maximize the likelihood of detecting coypu signs under favorable environmental conditions. Surveys spanned multiple seasons across the years, ensuring a balanced temporal and spatial coverage to account for potential variations in coypu detectability. Accessibility differences across habitats, such as open lakeshores, dense reed belts, or steep riverbanks, were addressed by allocating equal survey time and using targeted approaches tailored to the specific habitat type. For instance, in dense or less accessible habitats, surveys focused on likely entry points, such as trails or gaps in vegetation, to enhance detectability.

To ensure a thorough coverage of the study area, we employed visual encounter methods encompassing both on-foot and vehicular surveys. The coypu, known for its adaptability to human-modified landscapes and tolerance of human activity, particularly in areas with limited hunting pressure and minimal predation risk [66,67], can be reliably observed using these methods. On-foot surveys involved teams of two researchers traversing established trails and secondary roads near water bodies to scrutinize nearby vegetation and ensure a comprehensive coverage. For vehicle-based surveys, teams of two researchers scanned habitats along predefined road networks (including primary and secondary) at a speed of about 10–20 km/hr. Each survey began after sunrise and concluded three hours after sunset, depending on the size of the area to be covered, ensuring a thorough coverage of each cell. We used binoculars, GPS, cameras, and spotlights to facilitating accurate and detailed data collection across the study area.

To evaluate coypu presence (Figure 2a,d), we employed both direct (Figure 2b,c,e,f) and indirect observations of the species, i.e., tracks and signs, including slides, trails, droppings, and burrows (Figure 2g,h,i,j).

Identification of footprints was usually straightforward due to the distinctive features of the hindfoot, with four webbed toes adapted for swimming, and a unique free outer toe utilized for grooming [68]. Additionally, a tail drag mark left between footprints (Figure 2g) further aided identification. We placed special emphasis on droppings as presence indicators. Wherever feasible, we implemented a consistent 10 m wide belt search for coypu faeces across all water body categories [42,69,70]. These droppings, ranging from dark green to almost black, were cylindrical, approximately 5 cm long and 1.3 cm in diameter, often found floating in water, along trails, or at feeding sites (Figure 2i). When multiple individuals were observed, all individuals were recorded, irrespective of whether they were part of a group or solitary. To address potential pseudo-replication, indirect observations within a 340 m radius [71] were aggregated as a single record unless direct observations confirmed multiple individuals. Additionally, we sought evidence of other coypu activity, such as gnawing, clipping, or shredding indicative of plant consumption (Figure 2j).

2.3. Landscape Metrics

To obtain habitat feature data for further analysis, we employed ArcGIS 10.7 (ESRI Inc., Redlands, CA, USA), to extract and analyze habitat metrics. To achieve this, we created circular plots with a radius of 340 m, centered on each location of coypu presence. This radius was selected to approximate the home range actively utilized by each individual during the observation period [71], as an estimation of the area in which each animal is likely to spend the majority of its annual cycle. Given the unavailability of sex information for the observed individuals, we opted for a radius representing their average home range, derived from two methods (MCP and KDE95) across four seasons with values of 0.36 ± 0.6 km² (MCP, males), 0.23 ± 0.5 km² (MCP, females), 0.47 ± 1.5 km² (KDE95, males), and 0.39 ± 0.2 km² (KDE95, females) [71]. When these values are averaged, the combined mean home range size is 0.3625 km², corresponding to a radius of approximately 340 m.

Within each 340 m radius circle, we analyzed land cover type using the CORINE Land Cover (CLC) dataset with a spatial resolution of 100 m [72]. We extracted the total area and calculated the percentage cover of each land cover type within each circular plot. We also used the Tree Cover Density (TCD) subset of the COPERNICUS high-resolution layers [73] at a spatial resolution of 10 m to obtain type (deciduous, evergreen, and coniferous) and density data for arboreal vegetation. To characterize the vegetation types available to coypus within their estimated home ranges, we employed total productivity [74] as a key metric considering both the central tendency (mean) and variability (standard deviation) measures for each circular plot.

Furthermore, we extracted geomorphological features (altitude, mean inclination, aspect) from a Digital Elevation Model (DEM) of Greece with a 10 × 10 m resolution (provided by the Biodiversity Conservation Laboratory, University of the Aegean) and we evaluated proximity to various environmental factors potentially influencing coypu habitat selection. These included distance to the nearest water source, categorized by wetland type (lagoon, lake, river, canal, pond, marsh) and salinity (freshwater, brackish, saline), and distance from the nearest urban areas and roads as an indicator of the impact of human activities and infrastructure on coypu distribution. Finally, we determined whether each location fell within a Natura 2000 site to further contextualize the conservation significance of the observed habitat features [75].

2.4. Spatial Analysis of Coypu Presence

We employed a multi-faceted spatial analysis approach to look into the invasion patterns of the coypu within the study area. For broader spatial trends, we employed the Kernel Density Estimation (KDE), a non-parametric statistical technique [76]. Based on the geospatial data acquired through our field surveys, the KDE enabled us to estimate the underlying probability density of coypu records across the entire study area. To account for spatial clustering and ensure unbiased density estimates, we employed a kernel function that weighted the surrounding points based on their proximity to confirmed records. The selection of the bandwidth, a key parameter that influences the degree of smoothing in the density surface, was fixed at 340 m, corresponding to the coypu’s average home range radius [71]. This approach ensured that the generation of a highly reliable and informative density surface [77,78] accurately reflected the species’ spatial ecology and avoided over-smoothing, providing a reliable depiction of areas with heightened and diminished coypu activity.

However, due to the fact that these areas may not perfectly reflect the actual point of documentation because of the inherent mobility of the animals, especially in aquatic environments such as the ones in our study area, we implemented the 340 m snapping process; this is an approach particularly beneficial for mitigating potential inaccuracies arising from animal movement, a common challenge in ecological studies where features need to be aligned but suffer from slight misalignment due to data collection or processing errors [77,78,79].

To refine the distribution trends revealed by the KDE, we examined local-scale patterns using two local indicators of spatial association: the Getis-Ord G_i* statistic [80] and the Anselin local Moran’s I [81]. Our analysis commenced with the creation of a weighted point feature class, “ICOUNT”, representing the frequency of observed coypus. Utilizing the “Integrate” and “Collect Events” tools in the ArcGIS 10.7 Toolbox (ESRI Inc., Redlands, CA, USA), we ensured the spatial integrity of our dataset by snapping coincident features together [82], thus minimizing potential gaps or overlaps [83].

Thereafter, we used the Getis-Ord G_i* statistic, a robust tool for examining spatial autocorrelation [81], to analyze the spatial clustering of coypu records across diverse locations within the study area. This analysis allows for the identification of regions where coypu records surpass what would be expected by chance, unveiling the presence of hotspots and coldspots. The outcome of this analysis is manifested through z-scores and p-values, crucial for determining the significance of the observed presence clustering [80]. The z-score and p-value collectively convey the direction, strength, and significance of clustering. Positive z-scores with larger magnitudes indicate more pronounced clustering of high values, designating a hotspot. Conversely, negative z-scores with smaller magnitudes signify more intense clustering of low values, identifying a coldspot [83].

To enhance the verification and supplementation of the hotspot analysis, and to gain insights into localized clusters, spatial outliers, and regions with random spatial patterns of the coypu, we conducted cluster and outlier analyses using the Anselin local Moran’s I statistic [81]. This allowed us to detect both groupings and areas with abnormalities within the coypu records. Moreover, while hotspot analysis primarily identifies regions characterized by significantly higher or lower concentrations of records, cluster and outlier analysis broadens the scope by detecting both clusters of high and/or low values and individual data points that deviate significantly from the expected spatial pattern based on their neighbouring records. These outliers were further categorized based on their relative concentration compared to surrounding records.

2.5. Statistical Analysis

We conducted all statistical analyses using the SPSS software (v. 25.0, Armonk, NY, USA: IBM Corp.). Initially, to explore the dependence of coypu on environmental factors and landscape characteristics, chi-square tests were employed. Utilizing the results from the Getis-Ord G_i* statistic, we categorized coypu records into five categories based on the derived z-scores. The categorization included: (a) hotspots (z-score > 1.96), (b) moderate hotspots (1.65 < z-score < 1.96), (c) neutral spots (−1.65 < z-score < 1.65), (d) moderate coldspots (−1.96 < z-score < −1.65), and (e) coldspots (z-score < −1.96). Additionally, we calculated Cramér’s V as a measure of association to assess the strength and significance of the relationships between these categorical variables. This approach facilitated an examination of the potential dependence of the coypu on whether each occurrence was within a Natura 2000 area, the dominant land use category, and water body type. For continuous variables, including proximity to water, urban areas, and roads, we conducted a Welch’s analysis of variance, due to the unequal sample sizes of the five categories, followed by the Games–Howell test for multiple comparisons.

Subsequently, a binary logistic regression model was employed to analyze the factors influencing coypu presence within the study area. To facilitate a robust comparison between locations with coypu records and those without, we randomly generated pseudo-absences in ArcGIS 10.7 (ESRI Inc., Redlands, CA, USA). The use of pseudo-absences was rendered necessary by the challenges inherent in confirmatory absence data for a mobile species like the coypu, particularly in habitats where detection probability may vary due to vegetation density or limited accessibility. To ensure the ecological plausibility of pseudo-absences, we initially divided the study area into zones based on the KDE-derived presence density values. For each zone, a standard minimum distance between pseudo-absences and known presence records was established, considering home range size, adjusted for zone density level (larger distances in high-density zones). Thus, we maintained a minimum distance of 340 m from low-density and 680 m from high-density zones, as well as a minimum distance of 340 m between them. This ensured that pseudo-absences did not overlap with known records. The number of pseudo-absences matched precisely with documented records, ensuring a balanced dataset for analysis [84]. For these randomly selected points, we mirrored the data collection process employed for confirmed records, extracting habitat information, geomorphological features, and proximity to vital environmental factors within a 340 m radius.

Both actual presence data and generated pseudo-absences were integrated into a binary logistic regression model to assess the significance and contribution of various factors influencing the likelihood of coypu presence. Model accuracy was evaluated based on Nagelkerke’s R² [85] and the goodness of fit was assessed using the Hosmer–Lemeshow test. Discrimination ability was examined through a classification table of observed and predicted values, and the area under the Receiving Operating Characteristic (ROC) curve was employed to determine the ideal threshold point based on sensitivity and specificity for optimal prediction accuracy [86].

3. Results

3.1. Distribution of Coypu in Western–Central Greece

We identified a total of 133 coypu presence records within the study area. These comprised 89 direct sightings, 29 tracks, and 15 indirect signs, including five burrows, two droppings, seven roadkills, and one individual killed by a gunshot (Figure 3a).

Coypus were most frequently associated with permanently irrigated land, which accounted for 42.9% of the records, followed by wetland vegetation at 32.3%. Among water body categories, canals dominated, representing 51.6% of the recorded presences, while lagoons accounted for 18.85%. In terms of salinity, 113 records were located in freshwater habitats, seven in brackish habitats, and one in a saline environment. This preference for freshwater habitats aligns with their average proximity to water, standing at 213 ± 333 m, emphasizing a strong affinity for aquatic environments. The average distance from urban areas was 2686 ± 1715 m and the average distance from roads was 81 ± 112 m.

A significant proportion of the records (61.7) were within protected areas (Figure 3a). Specifically, 53 records were documented within GR2310001 (Acheloos delta, Mesolongi–Aitoliko lagoon, Evinos estuary, islands of Echinades and Petalas), 24 within GR2310009 (lakes Trichonida and Lysimachia), and five within GR2310007 (Lake Amvrakia). However, in some instances, individuals were seen outside the limits of Natura 2000 areas (n = 21), although not far from them: 18 were at a mean distance of 43 ± 23 m from the edge of GR2310001 and three were at a mean distance of 77 ± 12 m from the edge of GR2310007.

3.2. Invasion Spatial Patterns and Hotspot Areas of Coypu

The KDE analysis revealed distinct spatial patterns in coypu distribution, highlighting high-density areas that served as hotspots within the study area (Figure 3). These density surfaces highlighted localized areas of increased coypu activity, potentially reflecting habitat preferences or environmental constraints influencing the species’ spatial distribution.

The hotspot analysis identified 12 specific areas with z-scores greater than 1.96 (Figure 4), indicating statistically significant high-density clusters. Remarkably, six of these hotspots were concentrated in the south-eastern part of our study area, while the remaining six were situated in the central part, evenly distributed within the GR2310001 and GR2310009 areas. These hotspots encompassed a total of 36 coypu records, with z-scores ranging from 2.00 to 2.41 (p < 0.05).

In contrast, z-scores falling within the range of 1.65 to 1.96 (n = 2) indicated a moderate level of significance in the context of coypu clustering. These moderate hotspots were in GR2310001 (one) and in GR2310009 (one). Neutral areas, comprising 29 clusters (74 records) with z-scores from −1.65 to 1.65, indicate a lack of significant spatial clustering (p < 0.1). These neutral areas were distributed throughout the study area, suggesting that the observed spatial patterns did not significantly deviate from the expected pattern under spatial randomness.

Two moderate coldspots (three records), with z-scores from −1.96 to −1.65 (p < 0.1), were both in the south-eastern part of the study area, outside any PA (Figure 4). Lastly, 13 coldspots, positioned amidst the moderate cold- and neutral spots, were concentrated in the south-western part of the study area. With a negative z-score lower than −1.96 (Figure 4), this specific area suggests a spatial clustering of coypus below what would be anticipated by chance. These coldspots involved a total of 16 coypu records, displaying z-scores ranging from −2.00 to −2.99 (p < 0.05).

3.3. Identification of Significant Clusters of Coypu Distribution

Anselin local Moran’s I confirmed the significant hotspots and coldspots identified by the Getis Ord G_i* statistic, by identifying significant clusters based on z-scores and p-values (

Table 1. Anselin local Moran’s I statistic detailing cluster types, the number of coypus in each cluster along with corresponding mean z-scores and mean p-values for each location.

Cluster Type	Coypu (n)	Z-Score (Mean)	p-Value (Mean)
High–high	78	4.000 ± 1.008	0.013 ± 0.008
Low–low	42	4.571 ± 0.935	0.010 ± 0.001
High–low	0	-	-
Low–high	0	-	-
Not significant	13	0.393 ± 0.620	0.299 ± 0.988

).

We identified 22 locations categorized as high–high clusters, with 15 within GR2310001 and seven within GR2310009, displaying a positive z-score surpassing the critical value (z > 1.96, p < 0.05). This signifies a statistically significant concentration of coypus surrounded by areas with similarly high concentrations (Figure 5). Conversely, 27 low–low clusters, characterized by a negative z-score surpassing the critical value (z < −1.96, p < 0.05) and located outside any PA, indicated a significantly low concentration of coypus surrounded by areas with similarly low concentrations (Figure 5). The analysis revealed no instances of outliers falling within the categories of high–low and low–high clusters. However, non-significant clusters were dispersed in the north and north-east parts of the study area, encompassing locations with positive or negative z-scores below critical values.

3.4. Dependence of Coypu Distribution on Environmental and Landscape Factors

The chi-square analysis revealed a statistically significant association between the presence of coypus within a Natura 2000 area and the five categories of coypu distribution, derived from the hotspot analysis (χ² = 10.52, df = 4, p = 0.032). The dominant land cover types within the circular plot of coypu records also demonstrated a significant influence (χ² = 90.94, df = 28, p < 0.001). Water body types, encompassing lagoon, lakes, rivers, canals, ponds, and marshes exhibited a significant impact on the distribution of coypus (χ² = 41.87, df = 16, p < 0.001).

The outcomes of Cramér’s V revealed a moderate effect size (V = 0.247, p < 0.05) in the context of coypu categories and their dependence on whether records were situated within a PA. On the contrary, a high effect size (V = 0.820, p < 0.05) was identified concerning the influence of the dominant land use category. Additionally, a moderate effect size (V = 0.393, p < 0.05) was observed in the relationship with water body types.

The analysis of continuous variables, including proximity to water, urban areas, and roads, revealed significant differences among the coypu distribution categories. The Welch ANOVA revealed a significant effect of the distance from water [F (4, 13.261) = 5.402, p < 0.05], distance from urban areas [F (4, 11.371) = 54.412, p < 0.05], and distance from roads [F (4, 14.392) = 6.343, p < 0.05] on the five groups.

Subsequent post-hoc Games–Howell tests elucidated that the distance from water significantly differed between coypus within the hotspots (76.89 ± 131.48 m) and those in the moderate hotspot categories (4.48 ± 6.45 m; p = 0.019, 95% C.I. = 8.85, 135.87). Furthermore, coypus within neutral spots (124.84 ± 322.00) differed from those in moderate hotspots (p = 0.017, 95% C.I. = 15.31, 225.42). In terms of proximity to urban areas, coypus within hotspots (1863.07 ± 591.95 m) exhibited significant differences compared to those in neutral spots (2420.53 ± 1304.10 m; p = 0.022, 95% C.I. = 55.60, 1059.30) and coldspots (6200.39 ± 1062.17 m; p = 0.001, 95% C.I. = 3486.61, 5188.02). Notably, coypus in neutral spots differed from those in both moderate hotspots (1803.79 ± 235.14 m; p = 0.001, 95% C.I. = 40.11, 1193.36) and coldspots (p = 0.001, 95% C.I. = 2883.94, 4675.78). Additionally, coypus within coldspots showed distinctions from those in moderate coldspots (p = 0.001, 95% C.I. = 2986.74, 6297.32). Finally, regarding the distance from roads, coypus within the neutral spots (110.08 ± 129.01 m) differed from those in the hotspots (28.64 ± 42.87 m; p = 0.001, 95% C.I. = 35.27, 127.59), moderate hotspots (21.02 ± 27.34 m; p = 0.005, 95% C.I. = 25.66, 152.44), and moderate coldspots (31.60 ± 13.06 m; p = 0.001, 95% C.I. = 30.16, 126.18).

3.5. Modeling the Likelihood of Coypu Occurrence

The logistic regression analysis incorporated both observed coypu records and randomly generated pseudo-absences (Figure 6).

The logistic regression model yielded statistical significance [χ² (2, N = 266) = 339.92, p < 0.001], indicating the influential role of two predictor variables: tree cover density (TCD) and standard deviation of productivity (

Table 2. The logistic regression model illustrating the likelihood of coypu presence. Β = logistic coefficient; S.E. = standard error of estimate; Wald = Wald chi-square; df = degree of freedom; p-value = significance.

Predictor	Β	S.E.	Wald’s χ2	df	p-Value
TCD	−15.093	5.998	6.332	1	0.012
Productivity (std)	0.014	0.004	14.7	1	0.001
Constant	−2.476	1.075	5.304	1	0.021

). The direction and strength of these variables’ influence on coypu presence are reflected in their logistic coefficients (Β). For TCD, the negative coefficient indicates a strong negative association with coypu presence. As the amount of tree cover increases, the log-odds of coypu presence significantly decrease. Conversely, the positive coefficient for productivity (Β = 0.014) suggests a positive association, indicating that higher productivity (std) levels correspond to a greater probability of coypu presence.

Achieving an overall classification accuracy of 98.1%, the model demonstrated specific accuracies of 98.5% for presences and 97% for absences. Moreover, it attained a high Area Under the Curve (AUC) value of 0.998 (S.E. = 0.001, 95% CI 0.996–1.000, p < 0.001), signifying an excellent predictive performance. The Nagelkerke R² accounted for 96.2% of the total variance, highlighting the model’s high explanatory power. The Hosmer–Lemeshow goodness-of-fit test yielded a satisfactory result (Hosmer–Lemeshow = 0.201, p > 0.05), indicating that the model fits the data at an acceptable level.

4. Discussion

Our study provides a detailed investigation into the distribution and spatial patterns of Myocastor coypus in western–central Greece, a region with limited prior documentation of this invasive species. By employing a grid-based field survey approach and subsequent spatial analyses, we sought to identify the current extent of coypu presence, highlight key habitat associations, and pinpoint areas of potential range expansion. This approach addresses critical gaps in understanding the local invasion dynamics of the coypu, a species known for its significant ecological and economic impacts [36]. The findings of this study are intended to inform management and conservation efforts by offering baseline data to guide monitoring programs and control measures. After all, the issue of coypu invasion has largely escaped the attention of conservation authorities, and the extent of the problem remains poorly understood, with only a limited number of studies addressing this concern [59,64].

Western–central Greece was selected for this study due to a substantial knowledge gap regarding coypu distribution in this region, particularly compared to northern Greece, where the species’ presence has been better documented. Reports from citizen science platforms and local observations suggest a potential southward expansion, emphasizing the need for systematic investigation. Although verified records have historically been confined to the southern shores of the Amvrakikos Gulf, unverified sightings of individual coypus in the Peloponnese, including near Patra and within the Kotychi–Strofylia Wetlands National Park (pers. obs.), suggest the potential for further range extension into the Peloponnese. While these anecdotal observations were not included in the current analysis, they underscore the importance of focusing on the western–central region as a critical area for understanding ongoing invasion processes.

From our dataset of 133 newly recorded coypu occurrences in western–central Greece, the species’ southward expansion appears to be following a trajectory from northern Greece into wetland-rich areas in the region. This progression aligns with the habitat preferences of coypus, as the wetlands in western–central Greece, characterized by diverse water body types, offer ideal conditions for colonization and proliferation. Similarly to observations in Italy and France [35,87,88], the ease of movement of the species in the area is facilitated by irrigation and drainage canals, rivers and interconnected ditches that span many kilometres and form a complex and extensive hydrological network. These waterways serve as vital conduits for coypu dispersal, playing a pivotal role in the species’ expansion, and provide access to supplementary foraging grounds, contributing to coypu proliferation within the broader landscape.

Accurately quantifying the true population size of the coypu in this region presented a significant challenge. The 133 coypu presence records identified over the four-year monitoring period likely represent a minimum estimate of distribution rather than an accurate population count. While each record corresponds to a distinct observation of direct sightings or indirect signs (e.g., tracks, burrows, and droppings), it is acknowledged that our approach does not account for individual movement patterns, seasonal variability, or potential overlapping home ranges. As a result, these records should be interpreted as spatial indicators of coypu presence rather than direct proxies for population size. Further studies employing methods such as individual marking and recapture or genetic sampling would be required to produce robust population estimates.

One key limitation contributing to this potential underestimation stems from the primarily nocturnal behavior of coypus [89]. By relying solely on daytime and dusk surveys (concluding three hours after sunset), we may have inadvertently underestimated active populations. However, the decision to focus on these time periods was guided by established activity patterns observed in both the coypu’s native and introduced ranges, such as Slovakia [90] and Argentina [91,92], as well as findings on other large rodents [93], including semiaquatic ones [94]. More than that, despite its wide extent, our study area has one of the lowest human population densities in Greece, hosting a population of 228,069, representing a mere 2.3% of the national population [95], and, thus, human disturbance to the coypu is minimal. This is reflected in the significant effect of the distance of coypu localities from urban areas and roads, evidenced in other parts of its distribution too [96]. Additionally, the single documented instance of an individual killed by a gunshot suggests a generally low threat level from humans, potentially allowing for daytime activity compared to areas with higher anthropogenic pressure. This aligns with established ecological principles where increased human disturbance is known to drive nocturnal behaviour in various mammalian species [97].

Furthermore, there is no documented evidence of predation on coypus by large wild predators in our study area or the entire country. Although the Eurasian wolf (Canis lupus) is present in the area [98] and their ranges overlap, unlike regions like Italy [99], there are no recorded instances of wolf predation on coypus. Other potential predators in the area include mainly the red fox (Vulpes vulpes), the stone marten (Martes foina), the marsh harrier (Circus aeruginosus), and the eagle owl (Bubo bubo), all of which could possible prey only on young or very young individuals but the aquatic habits of the coypu would keep even young individuals safe from such predators in most situations. Interestingly, the only documented predation case was observed in Lake Kerkini, northern Greece, where a livestock guardian dog preyed upon a coypu, highlighting the potential influence of non-wild predators on coypu populations in Greece [100]. Thus, the absence, as far as we know, of significant predation pressure may contribute to the coypu’s relatively relaxed behavioral patterns in our study area compared to regions with higher predator densities. This aligns with broader findings on rodent responses to predation risk, which can drive behavioral adaptations and ecological shifts under selective pressure [101,102,103]. Such dynamics underscore the importance of predation—or its absence—in shaping coypu distribution that could have further contributed to a less pronounced nocturnal behaviour compared to other regions.

The high number of new coypu records within the study area strongly suggests their firm establishment in western–central Greece. This can likely be attributed to several environmental and landscape factors, including water body types, land cover types, and protected areas, all of which were found to significantly affect the probability of presence. The complex landscape of the study area, encompassing diverse ecological characteristics like marshes, lakes, rivers, and agricultural land [104], provides a mosaic of suitable habitats and potential dispersal routes for various species, including the coypu. Wetlands, recognized as ecosystems harboring a diverse array of habitats, flora, and fauna [105], acknowledged for their high nature and ecosystem services value [106], might offer suitable conditions for coypu establishment due to the presence of freshwater resources and potentially abundant food sources [39]. Simultaneously, the expansive terrestrial agricultural areas adjoining these wetlands play a crucial role toward biodiversity conservation [107], providing essential habitats for a diverse range of faunal species, offering the coypu and other wildlife ample opportunities for activities such as foraging, nesting, and breeding [108,109]. Additionally, the absence of hunting pressure on the species in Greece, coupled with the lack of natural predators and its high reproductive potential, producing multiple litters annually [41,110,111], is likely to facilitate a rapid population increase [36], mirroring the situation observed in Italy [112]. This trend is particularly concerning in western–central Greece, which harbors the third highest number of threatened species per administrative region (n = 98) in the country [113]. The presence of the coypu in this ecologically sensitive area may exacerbate challenges for native species conservation, especially for those reliant on or closely associated with the reedbed habitats that the coypu favors [114].

In light of these considerations, the analysis of environmental and landscape factors predicting coypu presence provided valuable insights, demonstrating the utility of a Binary Logistic Regression (BLR) model. The BLR model exhibited high effectiveness, with a discriminatory performance of 98.1% and a notable informativeness, explaining 96.2% of the variance in coypu presence. This suggests that the coypu’s presence can be reliably determined by considering variables such as TCD and the standard deviation of productivity. The negative association between TCD and coypu presence suggests that the coypu is less likely to occur in areas with a higher tree cover density. This could be because coypus prefer more open habitats with less dense vegetation [115], as dense tree cover may limit ground-level food sources as well as limiting access to suitable food sources or preferred habitats. Additionally, a high tree cover density may provide cover for potential predators of the coypu, leading them to avoid such areas. On the other hand, the positive association with the standard deviation of productivity suggests that the coypu is more likely to occur in areas with greater variability in productivity. This could indicate that coypus are attracted to areas with a diverse range of food sources [69,116,117] or habitat types [45], both of which may provide them with greater foraging opportunities and habitat complexity. A higher variability in productivity may also reflect more dynamic and heterogeneous ecosystems, which could provide coypus with a range of ecological niches and resources to exploit.

The spatial distribution patterns of the coypu records, as extracted by the KDE and the Getis-Ord Gi*, provided important information on the hotspots and coldspots for the species in the area. The density surfaces created through the KDE, despite highlighting the existence of clusters with an abundance of individuals in five regions of our study area (Figure 3), could not definitively ascertain if these clusters resulted from random chance or underlying ecological processes [118]. Simply visualizing patterns, while informative, is insufficient for confirming statistically significant hotspots of species occurrence [119]. Instead of relying solely on the KDE, the Getis-Ord Gi* statistic provided a robust assessment of spatial clustering patterns, allowing us to differentiate statistically significant hotspots from areas with a random distribution [80]. The identified coypu hotspots, in agreement with the KDE results, emphasize the central and the south-eastern part of western–central Greece as critical areas necessitating urgent mitigation measures, substantiated by statistically validated clustering patterns [120].

It is noteworthy that all identified hotspots were in lowland agricultural landscapes with irrigation systems. These areas, often adjacent to extensive wetlands with irrigation and drainage canals [121], likely offer a combination of easily accessible food resources (e.g., annual crops and meadows) [112] and suitable refuges (e.g., Phragmites sp., Juncus sp.) for the coypu. This observation aligns with prior research indicating that habitat features such as abundant aquatic vegetation, wetlands, and agricultural crops can contribute to high coypu densities [39]. For instance, the six hotspots near the eastern side of the Kleisova Lagoon (GR2310001) suggest an association with a marsh rich in aquatic vegetation, canals, streams, crops, and salt flats. Similarly, the central region, particularly the area between lakes Trichonida and Lysimacheia (GR2310009), harbors another six significant hotspots. The presence of rich aquatic vegetation and surrounding crops likely contributes to the high coypu density in this area.

Conversely, the lowland area near the Acheloos River mouth, despite exhibiting characteristics seemingly favorable for coypu presence, presents a conundrum. This predominantly rural landscape boasts a combination of tree and annual cultivations, including permanently flooded rice fields for most of the year, meadows, uncultivated areas, extensive agricultural canal systems, and swamps on both sides of the river. While these features might suggest a suitable habitat, the Getis-Ord Gi* statistic identified 13 coldspots in this region. This apparent contradiction highlights the complicated relationship between coypu presence and habitat features. Studies suggest that the coypu primarily consumes a wide variety of aquatic vegetation [122,123] and only resorts to terrestrial options when aquatic resources are scarce [42,116]. While rice fields might offer a partial substitute for natural aquatic habitats during flooded stages, their suitability might decline during post-harvest drying periods. This seasonal variation in habitat quality could contribute to fluctuations in coypu presence, potentially explaining the formation of cold spots during periods of reduced food availability. Furthermore, in its native range, the coypu may not readily utilize rice as a food source, and rice paddies might even serve as a buffer zone preventing them from invading other crops [124].

Finally, the absence of high–low and low–high clusters suggests a clear spatial segregation between areas of high and low coypu presence probability, reinforcing the notion of distinct habitat suitability within the study area. Furthermore, the identification of high–high clusters within designated PAs highlights the importance of implementing targeted coypu management strategies. Continued effective management practices, alongside potential population control measures, are crucial to maintain suitable habitat for native species and mitigate the negative ecological impacts associated with coypu populations. Conversely, low–low clusters outside PAs suggest areas of potentially lower habitat suitability. This information can be used to prioritise management efforts and minimize the risk of further expansion.

5. Conclusions

Our study on the coypu’s invasion dynamics in western–central Greece has contributed insights into the species’ current distribution and habitat preferences. Through extensive field surveys and spatial analyses, we document new records of the species and identify localized hotspots and coldspots within the study area. By integrating spatial statistical techniques and modeling approaches, we elucidate key environmental and landscape factors driving coypu distribution patterns.

The identification of areas with higher coypu records, particularly in lowland agricultural landscapes, wetlands, and interconnected hydrological networks, underscores the importance of these habitats in supporting coypu populations. Conversely, the identification of areas with lower coypu records, despite seemingly favorable habitat features, such as rice fields, highlights the complex interplay of environmental variables influencing coypu occurrence.

These findings emphasize the need to focus management efforts on areas with lower tree cover density near wetlands. Integrating these findings with sustainable land management practices, prioritizing research on predator–prey relationships, and quantifying the coypu’s impact on native species are critical next steps. Building on this knowledge base, we can design an evidence-based coypu management strategy that addresses the ecological challenges posed by this invasive species while contributing to biodiversity conservation and ecosystem stability.