Aquatic invasive species (AIS) are amongst the greatest threats to native aquatic biodiversity. These introduced species often thrive in human-altered environments and spread through human-mediated pathways to invade new watersheds. Corbicula fluminea is a freshwater bivalve native to southeastern Asia first introduced in North America in Seattle, WA, in 1938 and has spread to nearly every major watershed in the southeastern United States. In the present study, we use an information theoretic framework to compare landscape and stream habitat variables associated with C. fluminea presence across five HUC10 watersheds in the upper Savannah River basin of South Carolina and Georgia, USA. Predictive models included landscape-level and site-level habitat variables associated with agricultural, developed, and forested landscapes. Models with variables associated with forested and developed landscapes were the top performing models based on AICc values. In top performing models C. fluminea presence was positively correlated with increased stream width, but negatively correlated with substrates dominated by cobble. Lower performing models highlight positive correlations with the presence of upstream reservoirs and increased developed landscape surrounding the site. Identification of habitat and landscape correlates with invasive species presence may lead to more efficient introduction monitoring efforts for conservation managers.

Freshwater golden clam, basket clam, land-use change, early detection, ecosystem vulnerability, habitat suitability, passive dispersal, aquatic hitch-hikers

Less than 1% of the Earth is covered by freshwater ecosystems, yet these environments contain nearly 10% of all recognized animal species (Balian et al. 2008). These systems are critically important to humans for water supply, recreation and tourism, flood control, and food production. The greatest threat to freshwater ecosystems is habitat degradation, specifically the altering of flow patterns, pollution from runoff, and land-use change (Sala et al. 2000). Habitat degradation may also affect freshwater ecosystems indirectly through the facilitation of species introductions (Strayer 2010). Consequently, the introduction and proliferation of invasive species may displace native species and can disrupt ecosystem function in the invaded habitat (Strayer 2010). Human activities, specifically land-use changes from agriculture, urbanization, and the creation of reservoirs have led to an increase in habitat fragmentation and spread of invasive species, further endangering freshwater ecosystems (Vörösmarty et al. 2010).

The Basket Clam (Corbicula fluminea Müller, 1774) is a freshwater bivalve native to southeast Asia and introduced to North America, South America, Europe, and some countries of North Africa (Clavero et al. 2012; Crespo et al. 2015). In North America, the species first appeared in Seattle, WA in 1938, and quickly expanded across the country. It can now be found in 48 of the 50 United States of America (USGS 2025). Global dispersal is likely facilitated by the building of reservoirs as well as activities related to trade, including global shipping and the construction of shipping canals (Karatayev et al. 2007), the pet trade and use as a food resource ( Ferreira-Rodríguez et al. 2019). Once established in a new region C. fluminea may spread to new waterbodies through attachment on fishing gear and boat hulls, bait-bucket transfers, as well as passive dispersal (Ferreira-Rodríguez et al. 2019).

Aquatic invasive species (AIS) invade new habitats through vectors such as shipping ballast water (Holeck et al. 2004), aquarium releases, water gardens, deliberate stocking, bait buckets, and horticulture (Keller and Lodge 2007). Certain life history traits improve the likelihood that a species will be a successful AIS, including the following: high reproductive capacity and rates, smaller body size, broad physical habitat tolerances, early maturation, and asexual reproduction or self-fertilization (Kolar and Lodge 2001). C. fluminea possess many of the same life history traits as other AIS. C. fluminea can reproduce through self-fertilization (Strayer 1999) and can produce 35,000 offspring per breeding season (McMahon 2002), allowing them to establish new populations from only one individual. Outside of human intervention, C. fluminea can disperse passively with water current through the release of juveniles, surviving the gut biome of fish, and by attaching to the legs of waterfowl and shorebirds (McMahon 1982). Once released from an adult, larval C. fluminea have a short time window (100 hours) where they persist in the water column by “swimming” with an organ called the velum (Mackie and Claudi 2010). This stage allows juveniles to disperse greater distances. During this time juveniles can also attach to human or wildlife vectors, which is most likely the method for short-distance upstream dispersal (Pernecker et al. 2021). Further, C. fluminea have a high filtration rate, can feed on a variety of algae, and can efficiently incorporate nutrients into somatic and reproductive growth (McMahon 2002). These dispersal and foraging characteristics of C. fluminea likely improve their success when invading new freshwater ecosystems (Pigneur et al. 2012).

Early detection and prevention of invasive species are more cost-effective solutions than removal or sustained management (Coughlan et al. 2020). Surveillance of aquatic environments for invasive species often lags behind the establishment of new populations (Beric and MacIsaac 2015), but could be improved through a better understanding of habitat suitability of the invader. Across their range, C. fluminea are considered habitat generalists, occurring in a wide range of lentic and lotic conditions (Karatayev et al. 2005), and across various substrate classes (Karatayev et al. 2003; Schmidlin and Baur 2007; Kelley et al. 2022). In the southeastern United States, greater C. fluminea densities have been linked to agricultural land use, including the amount of agriculture in the watershed, increased water temperature, and increased nitrogen pollution (Ferreira-Rodríguez et al. 2022).

Our objective in this study is to understand the habitat and landscape variables that affect C. fluminea dispersal and presence in the upper Savannah River watershed in Georgia and South Carolina. Based on previous literature, we expect C. fluminea presence to be positively associated with the amount of agricultural land cover (Ferreira-Rodríguez et al. 2022) and the presence of reservoirs upstream. C. fluminea have been associated with reservoirs receiving recreational pressure and seen in higher abundances downstream of dams; therefore, we expect to see the same association in our study (Karatayev et al. 2005; Robb-Chavez et al. 2022). Understanding the habitat and landscape variables that affect C. fluminea dispersal may allow more effective monitoring and prioritization of habitats for preventative measures.

Corbicula fluminea are an AIS currently found throughout most major watersheds in the United States (Karatayev et al. 2007, USGS 2025), and whose effects on native biota are not fully understood (Haag et al. 2021). Often considered a habitat generalist (Kelley et al. 2022), C. fluminea may be associated with varying habitat correlates depending on the system they are invading. Nevertheless, understanding relationships with habitat and spatial variables could aid in targeted management or monitoring prior to the establishment of new populations. To better understand the distribution of C. fluminea in the upper Savannah River watershed of South Carolina and Georgia (USA), we used an information theoretic approach to evaluate multiple competing hypotheses of landscape use (agricultural, forest, developed) and site level habitat to best predict C. fluminea presence. We found C. fluminea presence was predicted best by increasing stream width, and negatively associated with an increase in cobble as the dominant substrate.

It is important to note that our study only evaluated presence of C. fluminea, but not abundance or density. We believe that presence only data can be helpful in defining suitable habitat for an invader, but may be improved upon to evaluate optimal habitat through the inclusion of quantitative estimates of density or abundance. We also did not evaluate detection probability in our study, and therefore cannot be certain that a recorded absence of C. fluminea is representative of true absence or imperfect detection. We used a rapid assessment protocol to maximize the number of sites that can be surveyed within a day, using a search rate of 10 m² per minute (Sterrett et al. 2018). Others have proposed that an effective search rate for freshwater bivalves in low densities is closer to 1–2 m² per minute (Strayer and Smith 2003; Smith 2006). While it is possible that the rapid search rate may have added false negatives at sites, C. fluminea often reach very high densities (e.g., 1.7–132 clams/m²; Kelley et al. 2022) and thus the likelihood of detecting at least one individual at these densities is high.

Our findings suggest that stream size was the most important factor associated with the presence of C. fluminea. As stream size increased, the probability of C. fluminea presence also increased (Fig. 2). The distribution and abundance of aquatic species is often related to increased stream size due to a greater diversity of habitat and available resources found in larger rivers or streams (Schlosser 1982; Vaughn and Taylor 2000). But in this case, the relationship could be more complex. Stream size was also highly correlated with larger catchment areas (Pearson r = 0.9, Suppl. material 1: fig S1). As catchments increase in size, there may be a greater number of accidental or deliberate introductions of invasive species through human activity simply due to greater area (Kolar and Lodge 2001). Once established, C. fluminea disperses downstream (Pernecker et al. 2021), aided by its free-swimming juvenile stage that lasts for approximately 100 hours (Mackie and Claudi 2010), sufficient time for larvae to reach the outflow of reservoirs and be carried further downstream. Therefore, it is possible that the stream width relationship in our study indicates a greater probability of C. fluminea being introduced higher in the watershed.

In our best explanatory model, we found a negative association between C. fluminea presence and increasing proportion of cobble as the dominant substrate (Fig. 2), which may be indicative of smaller, headwater streams in forested watersheds. Successful aquatic invaders may tolerate a wide range of habitats, and C. fluminea can withstand a range of flow conditions and substrate types in lentic and lotic systems. Previous studies have indicated increased presence and abundance in slow-moving sandy rivers (Schmidlin and Baur 2007), rivers with larger substrate (Kelley et al. 2022), and in reservoirs (Karatayev et al. 2003; Patrick et al. 2017). In a behavioral habitat preference experiment, Schmidlin and Baur (2007) found C. fluminea preferred fine sediments including sand and fine gravel. The behavioral preference corroborated observations of higher natural densities in fine sediments in the Rhine River (Schmidlin and Baur 2007). In our study, proportion of sandy substrates was also a significant and positive predictor of presence, but only in lower performing models (Agriculture 3, AICc = 56.930, AICcWT = 0.004; Suppl. material 1: table SS3). The lack of a strong positive relationship between presence and substrate may be indicative of C. fluminea’s ability to colonize multiple substrate types, but it is also difficult to compare our presence data with abundance data reported from other studies.

Freshwater ecosystems are often affected by changing land use (Sala et al. 2000), and land cover in the catchment of a site may affect in-stream habitat parameters (Jacobson et al. 2001). Land use can influence water flow and sediment loads, which in turn may impact benthic organisms, and critical habitat along and within the streams (Allan 2004). Ferreira-Rodríguez et al. (2022) found C. fluminea densities were positively associated with the amount of agricultural land cover in a watershed. We did not see any relationship between agricultural land use and C. fluminea presence, but it is possible that agricultural land use is a better predictor of abundance than of presence only. Conversely, we did see a significant negative relationship with the % forest, and a significant positive relationship with % developed land within a 3km buffer around a site. However, these relationships were reported in lesser performing models representing 3% (Forest 2) and 4% (Develop 1, Fig. 2) of AICcWT (Suppl. material 1: table SS3). The association of C. fluminea presence with surrounding land cover may indicate that habitat degradation promotes establishment and proliferation of AIS. For example, increased developed land use, and conversely declining forested land use adjacent to a stream may result in declining habitat quality (Allan et al. 1997), reduced native biodiversity (Strayer et al. 2003), and thereby foster an invasion. It is also feasible that increased developed landscape (or decreased forested landscape) may indicate increased introduction and propagule pressure from human aided dispersal. Rodríguez-Rey et al. (2023) found that anthropogenic variables related to dispersal (e.g., road density, distance to ports or boat ramps, and human population density) provide greater relative importance in predicting the presence of AIS in the United Kingdom. It is possible that the gradient response of forested vs. developed land use in our study represents a gradient of anthropogenic activity aiding in C. fluminea dispersal in the upper Savannah River basin.

Human activity within and between reservoirs also serves as a potential mechanism of overland dispersal for aquatic hitch-hikers. Reservoirs have been shown to act as stepping stones for AIS invasions by providing accessible environments for human activities serving as introduction vectors (Johnson and Carlton 1996; Johnson et al. 2008). As relatively new habitats, reservoirs generally have lower species biodiversity than their original lotic systems, resulting in open niches that invasive species can exploit (Wetzel 1990; Havel et al. 2015). Karatayev et al. (2005) related C. fluminea presence to landscape characteristics throughout the state of Texas, USA. The authors found that C. fluminea were disproportionally present in larger reservoirs compared to smaller ones due to the amount of human activity that these reservoirs received. We found a positive relationship between the presence of upstream reservoirs and the presence of C. fluminea (Fig. 3), though this was in a lesser performing model (Develop 1). The initial introduction of C. fluminea may have been facilitated by human recreational activities, including boating, swimming, and fishing, and may have spread to nearby reservoirs in the same manner (Havel et al. 2015), or through passive transport on the feet and feathers of aquatic birds (McMahon 2002). We did not, however, assess presence of C. fluminea in upstream reservoirs and we acknowledge that this pathway is speculative, yet a plausible explanation given support from the literature. The potential of reservoirs as stepping-stones of passive dispersal by C. fluminea warrants continued investigation.

Our findings suggest that some landscape variables related to forest and developed environments may play a role in determining C. fluminea presence; however, site level habitat characteristics including substrate and stream width showed clearer relationships to C. fluminea absence or presence. Differences in our results from those previously published may simply be a result of differences in habitat and landscape processes that predict presence vs. abundance. Whereas habitat and landscape correlates of abundance or density are more indicative of optimal or quality habitat, presence data may be helpful in identifying suitable habitat where an invasion may initially occur. Given support in the literature, future research on C. fluminea may benefit from further evaluation of upstream reservoirs and developed landscapes on clam presence and abundance.

ZS: research conceptualization, sample design & methodology, investigation and data collection, data analysis and interpretation, writing – original draft, writing – revisions; MB: research conceptualization, sample design & methodology, investigation and data collection, data analysis and interpretation, writing – revisions; MH: data analysis and interpretation, writing – revisions; BI: data analysis and interpretation, writing – revisions; PH: research conceptualization, sample design & methodology, investigation and data collection, data analysis and interpretation, writing – revisions, funding.

Ericah Beason and Morgan Kern of the South Carolina Department of Natural Resources (SCDNR) assisted with permits, site selection and access permission. We are indebted to Regina Mondibrown, Molly Martin, Amanda Van Buskirk, Jake Smith, and Hayley Robinson for assistance in data collection, and Robert Ratajczak for logistical support. The authors benefitted from early feedback from Allison Roy (Massachusetts Cooperative Fish & Wildlife Research Unit) and members of the Brook Floater Working Group. We also thank our two anonymous reviewers for their constructive feedback.

Funding for this project was provided by a United States Fish & Wildlife Service State Wildlife Grant to the SCDNR, and graduate assistant support from the University of Georgia Warnell School of Forestry and Natural Resources. The Georgia Cooperative Fish and Wildlife Research Unit is sponsored jointly by the Georgia Department of Natural Resources, the University of Georgia, the U.S. Fish and Wildlife Service, the U.S. Geological Survey, and the Wildlife Management Institute. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the United States Government.

All sampling activities were permitted by SCDNR under permit number F – 23 – 058, as well as written permission for stream access from private landowners and the United States Forest Service Francis Marion and Sumter National Forest.

All of the data that support the findings of this study are available in the main text or Supplementary material.