Introduction

Aquatic invasions can affect biodiversity and ecosystem services at different levels by outcompeting native species, disruptin trophic relationships, and—among others—altering nutrient dynamics (Havel et al. 2015; Bellard et al. 2016; Cuthbert et al. 2021). The extent of invasive non-native species introductions and the severity of their impacts depend on the connectivity of the aquatic system, the frequency and magnitude of novel introductions (i.e. propagule pressure and colonisation pressure; Lockwood et al. 2005; Briski et al. 2012), and the taxon considered, as well as the effectiveness of the consequently applied management actions (Ahmed et al. 2022). Likewise, distinguishing temporal and spatial dynamics within the invasion process is essential for informing decision-making, optimising current and future conservation interventions, as well as minimising the impacts of invasive species (Ricciardi 2007; Hulme 2009; Haubrock et al. 2022). A common method to estimate the potential spread of an invasive non-native species involves evaluating its niches in environmental and geographic contexts using species distribution modelling (SDM). These models can correlate species occurrences from the native and invaded ranges with environmental data to predict the environmental suitability and invasion potential in space and time (Soberón and Nakamura 2009; Araújo and Peterson 2012). Nevertheless, climatic modelling assumptions oversimplify the relationships between species and their environment and whether these relationships are affected by e.g. (i) genetic and phenotypic variations, (ii) dispersal features, (iii) biotic interactions and (iv) niche conservatism (Liu et al. 2022; Han et al. 2023). The niche conservatism hypothesis—a fundamental assumption on which SDM-based predictions have been developed—states that the niches of species remain conserved across different spatial and temporal scales (Soberón and Nakamura 2009; Wiens et al. 2010). Approaches for quantifying niche dynamics have also been developed to study the success of non-native invasive species in new environments (see a review of 434 species at Liu et al. 2020). While invasive non-native species often maintain their niches (Liu et al. 2022), niche evolution is a significant aspect of their invasiveness (Hui et al. 2023). This evolution can result from environmental selection in new ranges (Card et al. 2018), changes in ecological interactions like predation pressure (Tingley et al. 2014), and adaptations in morphology, physiology, and function (Macêdo et al. 2021). Understanding whether invasive non-native species retain their niches after successful establishments is crucial for more precise modelling of potential range expansions and for guiding effective management strategies.

Ecological and evolutionary changes, notably including biological invasions, are significantly altering and threatening biodiversity globally (Mooney and Cleland 2001). This highlights the need to investigate variations in landscapes affected by invasions and their diverse impacts on native species and their communities. In our study, we used SDMs to explore the ecological and biogeographical characteristics of Cercopagis pengoi Ostroumov, 1891, a small invertebrate with an adult body size of 1–3 mm that is listed as one of the ‘100 of the World’s Worst Invasive Alien Species’ (Luque et al. 2014). Native to the Caspian and Aral Seas, C. pengoi likely spread to the Great Lakes via ballast water from the Baltic Sea (MacIsaac et al. 1999; Cristescu et al. 2001). As a generalist predator, it competes with native zooplankton and meroplankton such as Leptodora kindtii Focke, 1844, and larval fish, leading to significant ecological and economic impacts (Ojaveer and Lumberg 1995; MacIsaac et al. 1999; Jacobs and MacIsaac 2007; Naumenko and Telesh 2019).

The basic premise of SDMs, namely niche conservatism, is key to understanding the invasion dynamics of C. pengoi. This species has successfully established populations in non-native areas, despite its low genetic diversity. Interestingly, it exhibits high phenotypic variation through cyclomorphosis—environmentally induced changes in cladoceran morphology (Gorokhova et al. 2000; Makarewicz et al. 2001; Litvinchuk and Telesh 2006). This suggests that physiological and morphological adaptability, a trait often seen in cladocerans (Gustafsson et al. 2005), might play a more crucial role in its invasion success than previously thought. Additionally, the limited geographic spread of C. pengoi over decades is noteworthy, as it suggests the potential to thrive in new environments. However, gathering reliable presence/absence data for C. pengoi, is challenging and nearly impossible to maintain globally. Thus, examining the current niches of C. pengoi in the Baltic Sea and North America, compared to its native Caspian region, offers an opportunity to uncover the patterns and biological processes shaping the niches of well-established invasive non-native species. Specifically, it allows us to investigate whether factors other than climatic shifts play a more significant role in determining their establishment and persistence, even at lower abundances (Spear et al., 2021).

Our first objective is to test whether the niche of the species shifts or remains unchanged after it establishes itself in non-native regions (Warren et al. 2008). We propose that niche conservation occurred from native to regional introductions in the Baltic Sea, while niche expansion plays a significant role in the successful establishment of populations overseas. The invasion success of C. pengoi was found to be more influenced by environmental filters rather than by biotic interactions (Golubkov et al. 2020), possibly due to the limited genetic variation between these areas (MacIsaac et al. 1999). By investigating the relative importance of niche conservatism versus niche shift, we aim to determine whether C. pengoi has spread to all potentially suitable habitats in the invaded ranges and, therefore, identify potential areas at risk for further spread. Specifically, we aim to assess whether the invasive ranges predict more similar potential areas compared to currently occupied ranges. The implications of our findings for the use of SDMs as tools for predicting new occurrences of non-native species are also briefly discussed.

Methods

Occurrence and environmental data

We gathered occurrence records of C. pengoi from (i) the Global Biodiversity Information Facility (GBIF 2021); (ii) the US Geological Survey’s Nonindigenous Aquatic Species Database (USGS-NAS; Fuller and Nielson 2015; Benson et al. 2023); and (iii) the Ocean Biodiversity Information System (OBIS; https://obis.org/taxon/234025); and (iv) literature-extracted data, which we searched by title, abstract, and keywords in Scopus and the Web of Science using the search strings: Cercopagis AND (“invas*” OR “non-native” OR “non-indigenous” OR “alien*” OR “introduc*”).

In total, we gathered 1,652 records of C. pengoi in the Ponto-Aralo-Caspian basin and two invasive ranges: ‘Invaded region 1’ (Inv1) in the Baltic Sea and ‘Invaded region 2’ (Inv2) in the Laurentian Great Lakes and Finger Lakes in the eastern North American region. We thinned data with a 10-km buffer around each coordinate using the spThin R library (Aiello-Lammens et al. 2015) to minimise unevenness in sampling efforts, geographical sampling biases, and spatial autocorrelation. As a result, we retained a subset of geographically unique occurrences (Native region = 11, Inv1 = 100, and Inv2 = 63) (Table S1 in Supplementary Information S1).

All bioclimatic variables for the current climatic conditions available from the WorldClim database (http://www.worldclim.org; Hijmans et al. 2005) were gathered at a 2.5 arc-min resolution for the analyses described in sections ‘Niche dynamics’ and ‘Projecting potential distribution in current scenarios’. We pre-selected variables to ensure the use of relevant proxies that better describe C. pengoi distribution and climatic niche as advised as one of the best practices for constructing species distribution and ecological niche models (Araújo et al. 2019). The selection of each variable was based on environmental characteristics that are key drivers of the population dynamics and reproductive patterns of C. pengoi. These factors may account for its successful invasion and its fitness within its native range, as discussed by Golubkov et al. (2020) and Litvinchuk (2021). The relevance of these variables is explained in detail in Table 1.

Table 1 Meaningful ecological variables for predicting environmental suitability and estimating niche of Cercopagis pengoi
Full size table

Niche dynamics

We estimated the amount of niche overlap, i.e. Schoener’s D (Schoener 1970; Warren et al. 2008), based on species abundance and the environmental conditions available for each population (Native region, Inv1, and Inv2), to test niche conservatism hypothesis. If the niche remains consistent throughout the invasion process, spatial models developed using data from the native range of C. pengoi should also indicate suitable conditions in the areas where the species has been introduced. We evaluated niche overlap with multivariate niche similarity tests, i.e. whether one niche is more similar to the others than would be expected by chance, given the available environmental background. Niche conservatism was inferred when the niche similarity test yielded significant values, suggesting that the niches are more similar than what random chance would predict. This means determining whether the climatic niche of one population can more accurately predict another population compared to niches randomly generated from a background region. To thoroughly interpret niche dynamics, we used the density of occurrences in environmental space. This approach helps estimating niche stability (i.e. the proportion of native niche conditions present in the non-native range), assesses niche expansion for insights into species adaptation (i.e. new environmental conditions encountered in the non-native range), and identifies environmental niche unfilling (i.e. the proportion of the native niche not occupied in the non-native range, as detailed by Guisan et al. 2014).

First, we generated an environmental space based on the principal components analysis (PCA) values calculated for the background and occurrence records. We then modelled the species density in the environmental grid, considering the observed occurrence density and the availability of the conditions in the background. We calculated the niche overlap and the partition of the non-overlapped niche among niche unfilling (i.e. niche space that is occupied in the native but unoccupied in the non-native range), expansion (i.e. niche space that is unoccupied in the native but occupied in the invaded range), and stability (i.e. the proportion of the non-native niche overlapping the native niche) (Warren et al. 2008; Broennimann et al. 2012; Guisan et al. 2014).

We extracted values for the four pre-selected bioclimatic variables (bio03, bio10, bio14, and bio15) at a resolution of 2.5 arc-min using the defined backgrounds (Fig. 1A). We randomised the occurrence records for both backgrounds. We generated a buffer of ~ 100 km around the occurrence records to further determine the background available conditions to further apply a PCA for all combined background environmental conditions and generate an environmental space (PCA-env; Broennimann et al. 2012). We calculated the occurrence density within each cell of the environmental space grid for the entire distribution range of the species. Finally, we estimated the environmental space using a smooth kernel density function that considers the geographical conditions available for the studied ranges (Broennimann et al. 2012). A total of 10,000 pseudo-absences were generated, and Schoener’s D was calculated 100 times to produce a null distribution of overlap scores (α = 0.05), which were then compared to the observed value (Warren et al. 2008).

Fig. 1
figure 1

Workflow for (A) species and (B) environmental data preparation, followed by the modelling procedures, is detailed. For niche dynamic analysis (C), a background derived from a Minimum Convex Polygon (MCP), constructed using occurrence records (Native region = 11, Inv1 = 100, and Inv2 = 63), was employed. In reciprocal distribution modelling (B), calibration was performed on one range with subsequent projection into another, and vice versa. Although the native range includes the Ponto-Aralo-Caspian basin and the Black Sea (denoted in blue as ‘Native’), it was not individually utilised in the training dataset for the native range in reciprocal modelling due to data insufficiency. Instead, a comprehensive dataset incorporating all records was employed. The invaded ranges are identified as the Baltic Sea and the Volga rivers in Russia (denoted in orange as ‘Invasive1’) and Northeastern North America in the Great Lakes (denoted in green as ‘Invasive2’)

Full size image

Projecting potential distribution in current scenarios

We used the same uncorrelated environmental variables used in the niche dynamics analysis in the reciprocal distribution modelling (Table 1). This approach allows identifying which geographic areas share similar climatic conditions between the calibrated and projected ranges. Results from the potential distribution of C. pengoi in both invaded ranges and the native range will uncover the areas not currently occupied (e.g. due to local extinctions or yet not reached) but with suitable climatic conditions for C. pengoi to prosper. This approach would also be suitable to infer the magnitude and direction of possible future range expansions. Our approach followed the workflow (Fig. 1A, C) where the models were calibrated in the native range and projected onto the two non-native ranges (Native region → Inv1, Native region → Inv2). Similarly, we calibrated models in the non-native ranges and projected them onto the native range and from one non-native range to the other (Inv1 → Native region, Inv1 → Inv2, and Inv2 → Inv1, Inv2 → Native region).

The modelling procedure was conducted using the sdm R package (Naimi and Araújo 2016). We used the machine-learning Maximum Entropy algorithm (Maxent; Phillips et al. 2006), one of the most utilised algorithms (Bradie and Leung 2017). This algorithm has good performance and accuracy, being flexible with absence data (Elith et al. 2011), a relevant feature when dealing with invasive species that are plausibly spreading below detection rates (Araújo and Peterson 2012). We generated randomly distributed background points in each calibration area (see Fig. 1) in a proportion of 1:10 (1 presence record to10 background points).

Following evidence-based recommendations, we fine-tuned process to determine the best Maxent settings for each calibration—projected area (All occurrences → Inv1 and Inv2, Inv1 → Inv2, and Inv2 → Inv1). For this procedure, we utilised the framework implemented in the ENMeval R package (Kass et al. 2021), where several configurations of features (L, LQ, LQH, and LQHP) and regularisation multipliers (1, 2, 3, 4, and 5) were tested to provide the best configuration for each model. We retained model settings that presented the lowest ∆AIC (Akaike Information Criterion; Akaike 1973). Models configurations are presented in Table S2 (Supplementary Information S2). We generated 20 replicates evaluated using bootstrapping or subsampling methods selecting 30% of random records for model evaluation. Model performance was estimated using (1) the area under the receiver operating curve (AUC; Fielding and Bell 1997), (2) true skills statistics (TSS; Allouche et al. 2006), and (3) the continuous Boyce index (CBI; Boyce et al. 2002; Hirzel et al. 2006). Lastly, we used the weighted average based on TSS to obtain the consensus model of the probability of occurrence. We considered the threshold that maximises sensitivity and specificity (max sp + se) as the cut-off value (Liu et al. 2013). All analyses were carried out in R language and environment version 4.3.0 (R Core Team 2023) with the ecospat R package (Di Cola et al. 2017).

Results

Niche comparison tests

The environmental niche spaces occupied by C. pengoi in its native and invaded ranges (Inv1 and Inv2) are represented in Fig. 2A–C. The first two PCA axes combined explained 77.1% of the environmental variation (52.1% for the first and 25.7% for the second axis; Fig. 2D). Isothermality (bio3) and precipitation seasonality (bio15) were the most important variables associated with the first axis (Fig. 2E) whereas the second axis was associated with the mean temperature of the warmest quarter (bio10) and precipitation of the driest month (bio14) (Fig. 2F). According to the PCA-env approach, native populations have more similar environmental conditions with Inv1 than with Inv2, which represents a subgroup of the native space (Fig. 3). In contrast, a low proportion of the environmental space in Inv2 is contained in the native space. The environmental space in North America (Inv2) has more adjacent space with the European range (Inv1), also shown by the environmental niche overlap between these ranges (Schoener’s D values; Table 1). The proportion of niche overlap exhibited between all three ranges is shown in Fig. 3, where the dashed lines represent the total background available in each region.

Fig. 2
figure 2

The first two axes of the principal component analysis (PCA), pertaining to the current distribution of Cercopagis pengoi in (A)–(B) invasive ranges and (C) native range, are presented. The PCA outputs for the preselected bioclimatic variables are shown in (D). The importance of variables in each PCA axis is detailed in (E) for the first axis and (F) for the second axis. The continuous line denotes 100% of the available environmental background, while the dashed line signifies the 50% most common conditions

Full size image
Fig. 3
figure 3

Overlaps in the realised climatic niches of Cercopagis pengoi across all considered ranges are depicted, with niches superimposed upon the available climatic background. The native range is represented in blue (Native); The non-native region Baltic Sea is represented in dark orange (Inv1) and the North American non-native region in green (Inv2). Continuous contours indicate the top 20% of density values, while dashed thin lines outline 100% of the available background in each respective region

Full size image

The similarity hypothesis test between Native and Inv1 (Table 2; p < 0.05) indicated that these niches were more similar to each other than to 100 randomly selected niches from the Native and Inv1 ranges (Broennimann et al. 2012). The observed overlap, based on Schoener’s D, is significantly higher than what would be expected to occur by chance. However, the similarity test results for Native-Inv2 and Inv1-Inv2 showed a non-significant similarity,suggesting that these niches are less similar than what would be expected by random chance, as detailed in Table 2.

Table 2 Multivariate niche comparisons and related metrics of niche dynamics
Full size table

Native and Inv2 had the lowest stability indicating dissimilar environmental conditions between both Ponto-Caspian and North American ranges (1 → 2, 28% and 2 → 1, 15%, Table 2). Furthermore, the higher degree of unfilling between Native region → Inv2 suggests that large portions of unoccupied environments remain available for C. pengoi in North America (unfilling = 85%). The niche dynamics analysis revealed a substantial expansion between the native range and the North American range (Native region → Inv2) at 72%, and between the Baltic and North American invasive ranges (Inv1 → Inv2) at 39%. In contrast, a smaller expansion of 16% was observed between the native range and the Baltic range (Native region → Inv1), as detailed in Table 2.

Reciprocal projections of SDM

Response curves showed that the distribution of C. pengoi is related to regions with less daily temperature variation compared to the annual range, with an optimal isothermality (bio3) slightly below 17%. Cercopagis pengoi displayed different suitability to average temperatures in its invaded ranges (around 16 °C in Inv1 and 20 °C in Inv2), however with similar temperature amplitudes (Fig. 4). Overall, C. pengoi demonstrates higher suitability in relatively stable and moderate hydrological regime, with populations in Inv2 having a slightly higher tolerance for precipitation seasonality compared to the Inv1.

Fig. 4
figure 4

Response curves for the environmental variables utilised in the potential distribution model for C. pengoi are detailed. These variables include isothermality (Bio3), mean temperature of warmest quarter (Bio10), precipitation of driest month (Bio14), and precipitation seasonality (Bio15)

Full size image

The reciprocal SDMs indicated an overall good performance considering all metrics utilised. Final models trained in all ranges (Native region + Inv1 + Inv2) had AUC 0.92 ± 0.01 (mean ± standard deviation), TSS 0.75 ± 0.03, and CBI 0.91 ± 0.05. Those trained in the Inv1 had AUC 0.92 ± 0.02, TSS 0.75 ± 0.06, and CBI 0.86 ± 0.06. Finally, when trained in Inv2, the model had AUC 0.97 ± 0.01, TSS 0.91 ± 0.04, and Boyce index (CBI) 0.66 ± 0.21. These results of the reciprocal modelling suggest that predictions calibrated in all ranges had an overall better performance than those calibrated only in the invaded regions (Fig. 5A).

Fig. 5
figure 5

Ensemble predictions for Cercopagis pengoi under the current scenario were derived using a weighted mean of models generated with MaxEnt. The binary suitability maps display the species’ predicted presence based on the maximum sensitivity plus specificity threshold (refer to the ‘Methods’ section for additional details). Colours indicate suitable areas, with red, yellow, and green representing varying degrees of suitability, while unsuitable areas are shown in grey. Maps correspond to predictions based on training data incorporating all occurrences (A), Inv1 (B), and Inv2 occurrences (C). Black dots mark the current distribution of C. pengoi

Full size image

The potential distribution of C. pengoi, as predicted by models calibrated in Inv1 (Fig. 5B), did not encompass areas currently found in the native range or within Inv2. On the other hand, models calibrated in Inv2 showed higher accuracy in predicting the occurrence of C. pengoi in its native range. These models also indicated more limited and geographically scattered regions in North America than those calibrated with all occurrences (Fig. 5C). The models trained in the invaded ranges displayed similar projected areas in Eurasia. They identified much larger geographic areas as climatically suitable for C. pengoi than its known current distribution, particularly in the Middle East and Russian water bodies. However, the risk of invasion is low in North Africa and Hudson Bay in Canada, as well as in most of the European territory and the Mediterranean Sea.

Discussion

In recent decades, a range of brackish and freshwater habitats in Europe and North America, including the Baltic Sea and the Laurentian Great Lakes, have experienced numerous invasions by species native to the Azov, Black and Caspian Seas. This phenomenon is known as the Ponto-Caspian invasion (Soto et al., 2023). Despite the environmental conditions and shipping frequencies in these regions, these invasions are far more prevalent than expected (Casties and Briski 2019). Euryhalinity, the ability to tolerate a wide range of salinities, is often investigated as a crucial invasiveness trait (Pauli and Briski 2018; Dobrzycka-Krahel et al. 2023). Nevertheless, climate conditions also play a significant role, often triggering sudden surges in invasive populations, specifically by accelerating the hatching of resting eggs (Spear et al. 2021). Our SDM specifically examined Cercopagis pengoi, a notorious high-ranking predator recognised by the IUCN as an impactful invasive species for over two decades (Luque et al. 2014). We sought to quantify the niche dynamics of C. pengoi, thereby enriching knowledge of niche instability as a mechanism behind plankton invasions. Importantly, our models identified regions potentially susceptible to the establishment of C. pengoi, emphasising the urgency of targeted preventive measures.

Niche dynamics

The debate on invasive population dynamics centres around two opposing ideas: one views invasive species as exhibiting niche conservatism, colonising environments akin to their native range, while the other suggests a niche shift occurs during the invasion process (Liu et al. 2020; 2022). In our study, we found support for partially rejecting the niche conservatism hypothesis during the invasion process of C. pengoi. However, invasive populations of C. pengoi demonstrated tendencies towards unfilling or evolution of their ecological niches, native environmental space was more similar to the non-native region of the Baltic Sea (Inv1) than would be expected by chance, a result consistent with niche conservatism. This suggests that the first non-native populations in the Baltic Sea likely originated from the native source range in the Ponto-Caspian region, also indicating a colonisation of populations affected by bottlenecks. This route is also supported by mitochondrial genetic analysis (MacIsaac et al. 1999) and by the low intraspecific genetic diversity commonly found in non-native populations (Gorokhova et al. 2000; Makarewicz et al. 2001; Litvinchuk and Telesh 2006). However, in this plausible initial invasion step, niche comparison analyses indicated that C. pengoi has only partially occupied its ancestral niche, suggesting that future invasions are anticipated.

Significant niche similarity was not observed when comparing the native space with the space of Inv2, nor between Inv1 and Inv2. Rather, a high degree of unfilling (i.e. conditions present in the native range but absent in the invasive range) and expansion (i.e. conditions unique to the invaded range) was evident between the native and North American populations (Inv2), as well as, to a lesser extent, between Inv1 and Inv2. This indicates greater stability, or the ability of the species to maintain its ecological niche over time (Warren et al. 2008; Broennimann et al. 2012). This implies that while the invasion process of Inv1 does not show a significant change in the environmental niche, the subsequent expansion into Inv2 could represent either the still-unfilled conditions in North America (indicating climatic non-equilibrium) or evolutionary processes occurring post-invasion in this region (Broennimann et al. 2012). The greater degree of unfilling observed between the Native region and Inv2, as compared to that between Inv1 and Inv2, might be attributed to dispersal filters rather than biotic resistance. This is because Inv1 is presumably a more intense source of propagules to Inv2, potentially sustaining ongoing propagule pressure. In other words, there is likely higher anthropogenic dispersal facilitated by the more active commercial route between the Baltic Sea and North America (Hudgins et al. 2023). Given that cladocerans rely on wind, drift, and other animals for the dispersal of their propagules (adult form or resting stages) (Moreno et al. 2019), it is predominantly human-assisted dispersal that facilitates their widespread distribution (Incagnone et al. 2015; Kotov et al. 2022).

In the study by Torres et al. (2018), a significant shift in the ecological niches of most freshwater invertebrates, including C. pengoi, was observed during their transition from native regions to potential new invasive spots in New Zealand. This shift in niche dynamics, as evidenced by the metrics in Table 2, could be attributed to rapid evolutionary changes or adaptations to novel environmental conditions. Such adaptations are particular features of cladocerans, which have short life cycles and well-known adaptive traits (LaMontagne and McCauley 2001; Gustafsson et al. 2005). However, the role of phenotypic plasticity in the establishment of C. pengoi also warrants consideration. Phenotypic plasticity, the ability of an organism to alter its phenotype in response to environmental variations, might be a key factor in the establishment of the species, as suggested in the invasion process of North American lakes by C. pengoi (MacIsaac et al. 1999). Indeed, if plasticity is a significant driver, the invasion pathway characterised by the sequence Native region > Inv1 > Inv2 might not strictly adhere to a pattern of pre-adaptation.

Northern regional and intercontinental range expansion

Suitable but not yet occupied environments can become time-bombs for sudden outbreaks of C. pengoi, as already shown for ecologically similar invasive cladocerans (e.g. Bythotrephes longimanus; Spear et al. 2021). Additionally, environmental non-equilibrium, as demonstrated here for C. pengoi, can undermine the accuracy of invasion risk assessments and thus deserves special attention from monitoring programs (Cutter et al., 2023; Maxson et al., 2023). These assessments can identify whether management biases or delays—attributed to the premise of the absence of invasive species due to non-detection—are plausible, given their ability to thrive in novel environments (Capinha et al. 2011; Gallardo et al. 2013; Torres et al. 2018; Macêdo et al. 2021). Curiously, C. pengoi populations are not established elsewhere in the world outside of the areas investigated in this study. In this sense, predicting the shifts in species ranges in response to environmental change is critical for developing timely conservation and mitigation strategies (i.e. anticipation or biocontrol; Dinis et al. 2020), and targeting costly invasive species (Cuthbert et al. 2021; Macêdo et al. 2022b), as inaction has been proven to be deleterious to global economies (Ahmed et al. 2022).

Our projections indicate improved model performances when based on both native and invasive ranges (Loo et al. 2007; Jiménez-Valverde et al. 2011; Macêdo et al. 2021). Invasive populations of C. pengoi may be adapted to different average temperatures while maintaining a consistent preference for limited daily temperature variability. Also, the responses to precipitation suggest that invasive populations have different drought resistance. These results suggest a degree of ecological plasticity in adapting to the mean thermal conditions of an area while maintaining a selective affinity for a narrow range of temperature fluctuations, which likely contributes to its invasive success in distinct yet climatically divergent regions.

Surpassing all occupancy restrictions, surveillance priority areas can be pointed out, aiming for more efficient control strategies for C. pengoi invasion. Starting from areas close to the source populations in North America and extending northward, e.g. the plains of Ontario and Quebec, Canada. Also, the predicted expansion southward into Illinois, possibly reaching even more southernmost locations in North America through natural downstream movements. For instance, Maxson et al. (2023) recently reported new records of C. pengoi in the Mississippi River Basin, which represents so far the southernmost record in the North American range. In Europe, we suggest higher efforts in Northern countries (Sweden, Finland, Denmark, and Norway), but primarily in Russian water bodies that have been invaded by zooplankton organisms through global trade (DiBacco et al. 2012; Casas-Monroy et al. 2015), including recent geographic range expansion of C. pengoi into Russian waterways (Lazareva 2019a, 2019b). The suitable areas for C. pengoi invasions indicated by our models are also targets for novel commercial ship routes in the Arctic (McCarthy et al. 2019), indicating areas where this cladoceran is currently absent (e.g. the Mediterranean Sea; Zenetos and Galanidi 2020). However, false negatives are possible as biodiversity in these regions has not been completely assessed, as new species are continuously being described (e.g. Alonso et al. 2021). Since methods of sterilisation of ballast water tanks can fail to reduce propagule pressure and instead trigger resting eggs hatching (Lin et al. 2020), implementing constant monitoring programs should be considered to mitigate both ecological and economic impacts, mainly to fish and fisheries (Macêdo et al. 2022b).

Conclusion

Species distribution modelling can provide insights into the mechanisms driving the invasion establishment of different populations of C. pengoi, beyond expectations of widespread niche conservatism of non-native species. Although relatively geographically restricted in invaded ranges, C. pengoi is not climatically constrained to the localities of its known current distribution, suggesting that more significant portions could be occupied if control policies weaken, or under novel environmental conditions that may trigger a population outbreak. Also, niches are neither similar nor stable between invasive ranges in Europe and North America, and introductions from these sources are possible or already ongoing but not tracked due to detection limitations. We suggest monitoring efforts should conduct samplings in bottom sediments due to resting egg bank formation for rapid detection and action, as well as developing protocols for ballast water sterilisation. In doing so, the invasion process of C. pengoi could be elucidated at local and regional scales, avoiding impacts on biodiversity and economic activities.