Understanding spatial distribution patterns is essential to management of invasive species. Aquatic invasive species can be notably challenging to detect due to the substantial effort required to locate them underwater. This limitation has resulted in a lack of timely distribution maps, particularly over vast regions, and hindered efforts to understand, forecast, and manage the proliferation of invasive bigheaded carps (Hypophthalmichthys spp.). Much of the Mississippi River basin, particularly the Lower Mississippi Alluvial Valley, provides access to a massive network of interconnected floodplain lakes. In the absence of lake-specific monitoring data on carp occurrence status, we used local expert knowledge, provided by fish managers interviewed virtually, in conjunction with Maximum Entropy (MaxEnt) modeling, to predict bigheaded carps distribution in relation to lake physical characteristics. We predicted widespread carp invasion in more than 60% of over one thousand floodplain lakes, with lake size, inundation, and proximity to rivers closely related to carp presence. The resultant distribution map may be imprecise given the swift proliferation of bigheaded carps and sparse monitoring data, but it offers a baseline upon which presence data and range can be compared. This assessment method is also a resource for identifying priority management and conservation areas and can serve as a first step in conservation planning.

Bighead carp, floodplain lakes, MaxEnt, silver carp, species distribution model

Invasive organisms can be a major threat to ecosystems by disrupting biodiversity and ecosystem functions (Sala et al. 2000). Competition for vital resources, such as food, water, and habitat, can be exacerbated when invasive species are present (Zwerschke et al. 2018). Human activities have enabled the spread of invasive species, both directly through intentional and unintentional translocations, and indirectly through habitat alterations (Ricciardi 2012). This is especially apparent in aquatic environments where artificial alterations in connectivity and flood regimes facilitate the dissemination and proliferation of invasive aquatic species and complicate the prediction of their spread pathways and distributions (Strayer 2010; Havel et al. 2015; Gallardo et al. 2016).

Understanding the geographical patterns of species distributions is key for the successful management of invasive species (Arim et al. 2006; Elith 2017). In aquatic systems, invaders can be rare, elusive, and nearly impossible to detect early during invasions as their low abundance makes monitoring colonization difficult (Jeliazkov et al. 2022). River fish are particularly difficult to sample due to high effort requirements (Dunn and Paukert 2020). In some cases, this is compounded by the lack of available methods for monitoring rare aquatic taxa (Lucas and Baras 2000; Wei et al. 2023). For example, physical surveillance may be impossible in remote riverine systems, and surveilling invasive species with molecular approaches (e.g., environmental DNA) is impractical over vast areas. This shortcoming could lead to an absence of timely distribution maps, which can hinder efforts to understand, predict, and control the spread of invasive aquatic species (Srivastava et al. 2019). Innovative new approaches are needed that are cost-effective and reliable, especially for species expected to expand swiftly and to have highly detrimental impacts to ecosystem integrity and services.

Bighead carp, Hypophthalmichthys nobilis (Richardson, 1845), and silver carp, H. molitrix (Valenciennes, 1884), collectively known as bigheaded carps, are invasive fish species in North America. They originated from East Asia and were moved to the Mississippi River basin in the 1970s for biocontrol in aquaculture operations. Likely due to flooding events, by the 1990s they had escaped captivity and began dispersing and reproducing across the Mississippi River basin (Kelly et al. 2011). It was not until 2007 that the first carp management plan was created (Conover et al. 2007; Brown et al. 2025). One goal of the management plan was to control the expansion of carp populations, mainly through monitoring coupled with rapid response programs. However, detecting invasive species such as carp when their abundance is low (i.e., at the invasion front) is difficult (Britton et al. 2011). The slow response and inadequate control methods have allowed bigheaded carps to rapidly expand in range and increase in abundance. Their expansion has elicited significant controversy and has prompted the natural resource conservation sector to allocate hundreds of millions of dollars to mitigate the observed and anticipated harm (Miranda 2023). Bigheaded carps are planktivorous filter feeders that consume substantial quantities of zooplankton and phytoplankton, potentially competing for food resources, directly or indirectly, with several native fish species at earlier life stages (Irons et al. 2007; Sampson et al. 2009; Minder and Pyron 2018; Harris et al. 2022). The significant disturbance they can cause to ecosystem processes has direct effects on native fish assemblages (DeBoer et al. 2018), altering fish assemblage composition and possibly biodiversity. The widespread success of bigheaded carps has been attributed to their extremely fast growth, early maturation, short generation time, highly adaptive nature, and access to eutrophic environments similar to those available in their native range (Miranda 2025).

Much of the Mississippi River basin, particularly the Lower Mississippi Alluvial Valley (LMAV), provides access to massive networks of interconnected floodplain lakes. Floodplain lakes are identified as the primary natural habitat for the development of juvenile bigheaded carps and for rapid growth to adulthood (Kolar et al. 2007; Chapman et al. 2016). At least 1,350 permanent floodplain lakes have been documented in the Lower Mississippi Alluvial Valley, in addition to many more ephemeral sloughs and swamplands (Baker et al. 1991; Miranda et al. 2021). These lakes vary in their level of connectivity to adjacent streams (Ahmad et al. 2024) and provide hypereutrophic environments that facilitate rapid growth of bigheaded carps (Killgore et al. 2024). Additionally, these lakes offer seasonal access to flowing water, which is important for bigheaded carp reproduction and dispersal. Furthermore, floodplain lakes in the LMAV provide important recreational, economic, and ecological value such as ecosystem services, fishing, hunting, and water supply for agriculture (Jenkins et al. 2010). Agricultural development has severely impacted these floodplain lakes, resulting in increased connections through the creation of drainage canals, redirected connections caused by levee systems development, reduced water quality, and elevated nutrients and turbidity (Miranda et al. 2014).

A key challenge in managing rapidly expanding invasive bigheaded carps in this extensive aquatic network is having information on their spatial distribution. Mapping the known and predicted locations of bigheaded carps is central to guiding effective management and assessing the effect of management actions (Gallien et al. 2010; Elith 2017). For example, predicting which lakes are suitable for carp to inhabit could allow biologists to prioritize early detection and monitoring efforts in vulnerable lakes. A predictive map could also offer a benchmark for comparing the spread of carp.

Conventional lake monitoring used for fish populations can be effective at relatively small scales but becomes expensive at scales as large as the hundreds of lakes scattered throughout the LMAV. One low-cost option for gathering information on carp distribution could be by collecting local expert knowledge, which is accumulated through personal observations by people working in a region. This type of information has been used for documenting species occurrences in previous studies (Anadón et al. 2009) and such data can be collected over large spatial scales with personal interviews, mail surveys, telephone surveys, and internet surveys at a relatively low cost (Meijaard et al. 2011; Barela et al. 2021). This option allows for the synthesis of data across many small-scale efforts to gain insight across large areas, at no additional cost to the small-scale programs. This approach has been successfully used to gather data about invasive species effects and distribution (Azzurro et al. 2019) and estimating species densities (Van der Hoeven et al. 2004). However, local expert knowledge has not, to our knowledge, been used to forecast the potential distribution of invasive fish. Therefore, we set out to use the local knowledge of fish managers to learn about the distribution patterns of bigheaded carps in floodplain lakes of the vast LMAV. Specific objectives included 1) use local expert knowledge and species distribution models to identify lakes that are suitable for bigheaded carps; 2) contrast the physical characteristics of lakes predicted to be suitable and unsuitable for carp; and 3) identify spatial patterns in predicted carp invasion.

A total of 26 interviews were completed with agency fish biologists from 13 fish management districts throughout the LMAV. We administered the questionnaire for 131 of the 1,086 lakes and recorded 64 carp presences (48.9% of sampled lakes) from lakes within the study area (Fig. 1; Suppl. material 1: table SS1). Interviews lasted 15 min to 2 h, depending on how many lakes were in each fish management district. The systematic random selection of lakes resulted in about half of the lakes chosen being in private lands, and therefore outside of fish management district jurisdiction. In most cases, respondents lacked specific information regarding carp presence in lakes within privately owned land, but respondents were often able to provide information about carp presence in adjacent public lakes. Additionally, twenty-four of these presence records came from supplementary information provided by agency biologists (i.e., datasets, reports).

Figure 1. 

(A) bigheaded carp (Hypophthalmichthys spp.) presences recorded during interviews with biologists and (B) lakes predicted suitable and unsuitable for carp presence in floodplain lakes of the Lower Mississippi Alluvial Valley.

The LASSO model implemented with logistic regression identified nine environmental variables suitable for inclusion as covariates in the MaxEnt model. These variables included lake area, lake length, probability depth, inundation index, agriculture cover, batture, disconnectedness, distance to river, and nearest neighbor (Table 1). Pearson’s product moment correlations among these covariates were generally low (i.e., r < 0.60; Suppl. material 2: fig. S1). The MaxEnt model had AUC-ROC values ranging from 0.78 to 0.88, TSS values ranging from 0.60 to 0.70, and Boyce index values ranging from -0.71 to 0.87 across each fold from the 5-fold cross validation. The AUC-ROC value was 0.87, TSS was 0.62, and Boyce index was 0.83 when the model was run using all data. Lake area, inundation index, and distance to river had the highest percent contribution to the model, and lake area, inundation index, and lake length had the highest permutation importance (Table 2). Smaller lakes that had high inundation potential and were closer to rivers were more likely to have carp present (Fig. 2).

Figure 2. 

Mean ± 95% confidence intervals lake attributes in relation to predicted bigheaded carp (Hypophthalmichthys spp.) status in 1,089 lakes of the Lower Mississippi Alluvial Valley, as of 2023.

Adding the maximum model sensitivity to the model specificity threshold resulted in a threshold of 0.24, which was used to transform continuous habitat suitability values to binary predicted presences (>0.24) and absences (<0.24). This transformation forecasted that roughly 62% of the lakes in the LMAV supported bigheaded carps (Fig. 1b). The environmental characteristics were different between the lakes likely to support carp and lakes not likely to support carp (PERMANOVA Pseudo-F = 21.3, P = 0.001). Each of the nine environmental factors varied in lakes that were predicted to be suitable for carp and those that were not. The most pronounced disparities reflected the MaxEnt model, which included lake area, lake length, inundation index, and distance from lake to major river, except that batture also emerged as a disparity (Fig. 2). Smaller lakes with high inundation potential that were closer to rivers were more likely to have characteristics suitable to carp invasion (Suppl. material 3: fig. S2).

We observed a geographic pattern in predicted carp habitat suitability across the LMAV. The Moran’s I index value was 0.12, revealing a positive and significant (P < 0.001) spatial autocorrelation, indicating that lakes that may support carp tend to cluster. In addition, nearest neighbor was moderately important in predicting suitability in the MaxEnt model. Clustering predominantly occurred along major rivers (Fig. 1). At the vast scale of the LMAV, no other clustering was evident.

Bigheaded carps are widespread and continue to expand their distribution in the Mississippi River basin, creating significant risks to the sustainability of native freshwater ecosystems (DeBoer et al. 2018; Miranda 2023). Our study has produced the first published map that documents the potential spread of bigheaded carp across the vast alluvial valley of the Lower Mississippi River. Although by definition pseudoabsences may or may not represent true absences, we are confident that our pseudoabsences are reasonably accurate because 1) carp invaded decades ago, allowing biologists and the public ample opportunities to detect the fish, and 2) bigheaded carps are large and conspicuous, especially silver carp, which frequently leap out of the water. Additionally, our model was developed using several methods recommended to generate more accurate predictions of suitability (Yackulic et al. 2013). These methods included collecting presence data using standardized sampling, considering how carp detection varies with environmental variables such as lake area, and providing all necessary information to evaluate results (e.g., response curves).

We acknowledge that while the map may be imprecise given the swift proliferation of bigheaded carps and sparse monitoring data, the map is still an informative management tool. Its value lies in revealing broad patterns, core areas, prioritizing surveillance and monitoring, guiding interventions, and supporting communication and planning. The map shows that this expansion can occur throughout this extensive floodplain given the lake characteristics available to support occurrence of carps in this region. Considering that these lakes offer bigheaded carp preferred habitats and that carps have access to a diversity of river systems to facilitate reproduction and dispersal, we anticipate that dispersal throughout the Mississippi basin is likely, barring obstacles (e.g., dams) or infrequent connections (Ahmad et al. 2024). Stafford (2024) observed several long-range dispersal events by silver carp in the LMAV, including a single fish traveling 1,210 km in less than 1 year. Ahmad et al. (2024) reported that most lakes in the LMAV became disconnected, with some disconnections lasting up to 20 consecutive years. This should be sufficient to allow local extinction, as carp are short-lived (i.e., 8–10 years, Killgore et al. 2024) and unable to reproduce in lentic systems. Barriers in the discharge channels of these lakes may impede future repopulation, and harvesting programs may accelerate extinction. Deployment of monitoring and management activities can be informed by the resulting distribution map.

In the absence of lake-specific monitoring data on carp occurrence status, we used local expert knowledge in conjunction with MaxEnt to predict bigheaded carp potential occurrences based on lake physical characteristics. We predicted widespread carp invasion throughout LMAV lakes with lake size, inundation potential, and proximity to rivers closely related to carp distribution in the LMAV. Most lakes likely to have carps were < 25 ha in size and were <5 km from a river. We suspect that these small lakes may have a higher likelihood of carp occurrence because their river proximity exacerbates the chance of recurring flooding that allows for carp dispersal and occupancy. A potential bias is that carp may be more visible in smaller lakes because these waters are easier to scan and have a higher proportion of littoral area, which is their preferred environment (Besson et al. 2024). Additionally, because many of the small lakes were predominantly positioned along the White River in Arkansas, a region targeted for carp removal, enhanced surveillance may have resulted in increased local awareness in this northwestern region of the LMAV.

The relationships between inundation index, distance to rivers, and predicted habitat suitability are intuitive. Inundation index is a measure of long-term fluctuation in lake volume. Predicted habitat suitability was generally positive, where lakes with higher inundation index were predicted to be more likely to be suitable for bigheaded carp. A high inundation index means it is more likely that a lake remains close to full volume over time. Bigheaded carps are tolerant of poor water-quality (Kolar et al. 2007), so they can persist in floodplain lakes as long as lakes do not go dry, thus potentially explaining the importance of inundation. The relationship between distance to river and predicted habitat suitability was negative, where lakes with shorter distances to rivers were predicted more likely to have habitat suitable for carp. Lakes that are closer to rivers have a higher probability of initial invasion simply because the carp have more opportunities to reach them. However, given bigheaded carp cannot reproduce within lentic waterbodies (Kolar et al. 2007), hydrologic connections likely enable carp to exit lakes, spawn in adjacent rivers, and re-invade either through the direct channel connections or connections made by over-bank flooding.

We determined a significant geographic pattern existed in predicted carp occurrence across the LMAV, with nearly two thirds of lakes predicted to be suitable for carp often clustering in the northwest LMAV along the White River. Similarly, nearest neighbor was moderately important in predicting lake suitability. Because bigheaded carp were first released near this region, they have had a longer period to invade. Moreover, fish scientists in this region may have developed more programs and competence for surveillance of bigheaded carp populations. Spatial clustering of predicted carp occurrences within the river system of initial introduction could foreshadow that bigheaded carps will eventually invade higher densities of lakes throughout the LMAV.

Using local expert knowledge to gather invasive fish presence data was cost-effective and relatively low effort, but there were challenges. First, local expert knowledge is based on personal observations, which can be influenced by individuals’ capacity to recall information and by the time elapsed since the waterbody was last visited. Moreover, respondents likely had varying tenures in their position, which also could have affected their knowledge of presence. We did not attempt to account for any tenure effects, as the relationship between tenure and knowledge, although presumably positive, was not documented. Second, because lakes were randomly selected, fish biologists were unable to provide information for roughly half of the lakes selected because they were on private land or otherwise not managed by their agency. While we could have obtained a larger sample size by collecting information only for lakes that biologists had knowledge about, it was important to select lakes at random to ensure that the sample of study lakes was representative of the breadth of lakes in the LMAV, reducing bias associated with agency management foci, and increasing the reliability of our results. Future studies may incorporate private landowners to obtain presence data for inaccessible lakes. Another drawback of using local expert knowledge is that it does not allow for the collection of absence data. Presence-absence data have been shown to be more reliable than presence-only data because presence-only runs the risk of not detecting fish that are actually present, such as where invasive species occur at low densities (Elith et al. 2020). Despite these challenges, its cost-effectiveness could make this method an excellent choice for large-scale work. The method can also be used to create new or to update distribution maps in other parts of the Mississippi basin, or to explore other elements of distribution, like size structure or temporal distribution.

The potential for widespread carp distribution identified in the LMAV highlights the need for rapid and aggressive management action. The map we produced offers a baseline for which data collection on presences and range expansion can be compared, and a resource for identifying priority management and conservation areas. Our results could be applied in many ways, but we see them used primarily as a first step in conservation planning. For instance, agencies may sample lakes predicted to not support carp to confirm their absence. They may then develop preventative measures against invasion such as barriers or other exclusionary constructions. Alternatively, agencies may target lakes predicted to support carp to confirm presence and then decide what management action would be most beneficial to minimize their effects.

Regardless of carp predicted status, continued monitoring through on-site surveys or local expert knowledge surveys such as ours are necessary to validate and update our estimates before and after conservation strategies are implemented. Modeling bigheaded carp distribution in the hundreds of LMAV oxbow lakes was an important step that can facilitate early detection and can guide on-site monitoring. This study offered a framework for an economical instrument that facilitates an initial assessment of large-scale invasive species monitoring.

We thank Neal Jackson and three anonymous reviewers for constructive evaluations. This publication is a contribution of the Forest and Wildlife Research Center at Mississippi State University. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. The survey described in this article was organized and implemented by the authors and was not conducted on behalf of the U.S. Geological Survey.

MP contributed to the research conceptualization, sample design and methodology, carried out data collection, performed data analysis and interpretation, ethics approval, and writing both the original draft and reviews and editing. LEM contributed to the research conceptualization, sample design and methodology, data collection, interpretation, ethics approval, funding provision and writing both the original draft and reviews and editing. MRB contributed to data analysis and interpretation, ethics approval, and review and editing. CGD contributed to sample design and methodology, data analysis and interpretation, ethics approval, and review and editing. LMB contributed to sample design and methodology, ethics approval, and review and editing. DR contributed to sample design and methodology, ethics approval, and review and editing.

Funding was provided by the U.S. Fish and Wildlife Service, Grant F21AP02996-00, to the Mississippi Department of Wildlife, Fisheries and Parks. The Mississippi Cooperative Fish and Wildlife Research Unit is jointly sponsored by the Mississippi Department of Wildlife, Fisheries and Parks; Mississippi State University; the U.S. Geological Survey; the U.S. Fish and Wildlife Service; and the Wildlife Management Institute.

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

Interview introduction script.

I am a graduate student at Mississippi State University. My thesis project is on bigheaded carp distribution patterns in the Lower Mississippi Alluvial Valley. Through a short survey I am gathering data on carp presence and lake characteristics that influence carp distribution. I will recommend management strategies based on the results.

The questions I ask you will be mostly closed ended. For each lake, I will read out the questions and record your responses. I will share a map of the lake to be sure we are talking about the same place. After each round of questions, I will ask if there are any adjacent lakes within a certain area, shown on the map, that you have information about. I will then administer the questionnaire for those lakes as well. You may ask for clarification at any time. I will repeat the questionnaire for as many lakes as we can within our allotted interview time. Do you have any questions before we begin?

Interview questionnaire.

Lake ID:

State and management district:

Effort/Detectability

1. What is the name of this lake?
2. How long has your agency been responsible for monitoring this lake?

• <1 year
• 1–5 years
• >5 years

1. How many times per year is this lake visited by a fisheries professional from your district?

• >2 per week
• 1 per week
• 1–3 per month
• 2–10 per year
• 1 per year
• 1+ in 5 years

Carp Presence

1. Have you seen silver or bighead carps or signs of them in this lake? If so, how often?

• never
• rarely (1–2 times)
• commonly (2+ times)

1. Have there ever been reports of silver or bighead carp in this lake from other citizens, anglers, landowners, or other fisheries professionals?

• yes
• no

1. If yes to 4 or 5, how long have the silver or bighead carp been in this lake?

• <2 years
• 2–5 years
• >5 years

1. In the last 5 years, have you noticed any changes in the fishery or fish community that you attribute to bigheaded carp?

• yes
• no

1. Are there any water control structures on the lake?

• yes
• no

1. If yes, where are the water control structure(s) on the lake located?

Current Management Efforts (to be asked at the end of the interview)

1. Are there currently any invasive carp management strategies being implemented in your district? If so, what strategies?
2. What percent of oxbow lakes have carp control strategies in place?
3. Have there been any complaints about Silver Carp or Bighead Carp from the public that use lakes in your district?