Introduction

The post-establishment population dynamics of invasive species are generally less well understood than the initial spread and establishment phases, which can be readily monitored via biodiversity surveys (direct sampling or eDNA) or citizen science observations (Larson et al. 2020; Callaghan et al. 2021). Some introduced species exhibit phenotypic or behavioural changes associated with the shift from colonisation to the post-establishment phase (Gutowsky and Fox 2012; Pietrek and González-Roglich 2015; Marchini et al. 2018). Developing an operational understanding of trends in population size, structure and distribution during this dynamic post-establishment phase often requires data collected over multiple generations. For species subject to ‘boom and bust’ cycles associated with environmental variability (e.g., large-scale climatic events), data may need to be collected for several decades after the initial population expansion before patterns in population dynamics and responses to environmental conditions become apparent (Simberloff and Gibbons 2004).

The logistic population curve is a well-recognised conceptualisation of how populations of invasive species establish and grow. It includes a rapid establishment and expansion period followed by a reduction in growth rate as the carrying capacity is approached (Hutchinson 1948; Gilpin and Ayala 1973; Sibly et al. 2005). Theoretically, the maximum density of a species is governed by the carrying capacity of the ecosystem, and it is often assumed that invasive species will continue to grow in number until they become dominant, resulting in exclusion or severe reductions of other species (Blackburn et al. 2019). The extent to which invasive species dominate ecosystems in the long-term is influenced by natural fluctuations as the population stabilises at the carrying capacity, while also being impacted by external factors such as responses to climatic variation and population control measures implemented for conservation purposes (Sakai et al. 2001). To develop effective invasive species management strategies, it is necessary to understand the population dynamics of the invasive species relative to other key species that inhabit the ecosystem.

One of the most successful and destructive invasive species in freshwater ecosystems is the Common carp Cyprinus carpio (henceforth carp). Despite being considered vulnerable to extinction in their natural range in the Black, Caspian and Aral Sea basins (Freyhof and Kottelat 2008), self-sustaining populations of carp have established in parts of Europe, Asia, Africa, North America, South America, Australia and Oceania (Koehn 2004; Lowe et al. 2004). In addition to providing an important food source for people in some regions (FAO 2022a, b), carp are a highly popular recreational angling species in some areas (Rapp et al. 2008), and both their importance for food security and angling has contributed to their geographical spread. However, their negative impacts on aquatic ecosystems outside of their natural range are well documented. For example, high densities of carp have been linked to decreases in macrophytes, increased water turbidity and nutrient concentrations, changes to fish and waterbird assemblages, and lower benthic invertebrate species richness (Kulhanek et al. 2011; Vilizzi et al. 2014; Bajer and Sorensen 2015; Bajer et al. 2016; Maceda-Veiga et al. 2017; Fanson et al. 2024). Carp are also thought to have major impacts on aquatic food webs by locking up basal energy resources (as carp biomass) that would otherwise be available to support the productivity of native fish populations and other food web components (Kopf et al. 2019b). The social impacts of carp are felt by communities through a loss of environmental quality and amenity, including reduced feelings of wellbeing and economic costs associated with the impacts of carp on traditional and recreational fishing, tourism and other activities (Wallis et al. 2009; Arlinghaus et al. 2017).

To mitigate the impacts of carp, authorities in many regions of the world undertake active management to reduce biomass via targeted eradication or control programs (Dalu et al. 2020; Yick et al. 2021; Piczak et al. 2023). Such activities include physical removal (e.g., Pinto et al. 2005; Yick et al. 2021), targeted commercial fishing (Koehn et al. 2000), exclusion screening (e.g., Hillyard et al. 2010), and research into options for biological control, including the production of “daughterless” carp to release into the wild (Thresher et al. 2014) and the release of the carp herpes virus CyHV-3 (NCCP 2022).

In Australia, carp populations are now established across the southeast temperate region and in limited regions of south-western Australia. Abundance and biomass estimates based on analyses of extensive survey data suggest there are ≈ 200 million individual carp (≈ 215,000 t) in Australia during average hydrological conditions (Stuart et al. 2021). The eastern region is estimated to account for 96% of the total carp biomass estimate in Australia, with perennial rivers of the Murray–Darling Basin (MDB) having the highest biomass density of up to 3,144 kg ha−1 (Harris and Gehrke 1997). Carp are considered so dominant that it is widely stated in the scientific literature and media that they contribute up to 90% of the fish biomass in MDB rivers (e.g., Koehn 2004; Pinto et al. 2005; BBC News 2016; Wahlquist 2017; McColl and Sunarto 2020; Pera et al. 2021; Mankad et al. 2022; Stuart et al. 2023). When traced back to its origin, however, this estimate appears to be based primarily on limited data from a small number of fish surveys conducted in the 1990s (Gehrke et al. 1995; Harris and Gehrke 1997; Koehn 2004) or from surveys conducted over limited spatial and temporal scales (Gilligan 2005). Recent time series analysis across the NSW MDB have reported large temporal fluctuations in relative abundance and biomass of carp, suggesting that the contribution of carp to total fish biomass may not be static over time (Crook et al. 2023).

The current study addresses the need for detailed, empirical analysis of available data to provide more comprehensive and contemporary estimates of the contribution of carp to total fish biomass across MDB rivers. We combine 30 years of standardised surveys of fish populations conducted in the New South Wales (NSW) portion of the MDB. Generalised linear mixed modelling is used to quantify the proportion of total fish biomass attributed to carp over spatial scales including the NSW MDB, water resource planning areas, and key rivers since the mid-1990s. The analysis also provides information regarding the post-establishment population dynamics of carp. Our findings are used to discuss the relevance of the revised estimates for the implementation of management and control options for this highly successful aquatic invader.

Methods

Data extraction and preparation

The state of NSW covers ≈ 57% of the MDB. Early fish surveys including the NSW Rivers Survey (Harris and Gehrke 1997), and Sustainable Rivers Audit (SRA; Davies et al. 2010), established protocols for standardised sampling of fish communities. These projects have conducted electrofishing surveys across the NSW MDB (and beyond) since the mid-1990s. Data for the current analysis were extracted from the NSW Department of Primary Industries (NSW DPI) Fisheries Freshwater Ecosystem Research (FER) database, encompassing data collected across 142 monitoring and research projects between 1994 and July 2023. The FER database includes catch (numbers for each species) and biological data (length, weights and health assessments) collected using a range of methods, including boat and backpack electrofishing, netting, angling and eDNA surveys. In total there were 14,110 sampling events (consisting of multiple samples) recorded in the database across 4,757 sites with > 2 million fish sampled from 219 taxa. To ensure consistency, only electrofishing data were used in the analyses (see below) and only that collected using standardised techniques (see Robinson et al. (2019) for details). Although there is some variation in the number of samples taken at each site among projects, boat electrofishing for most surveys usually consisted of 12 ‘shots’ of 90 s power on time while backpack electrofishing usually consisted of 8 ‘shots’ of 150 s of power on time. Other sampling methods are used in the NSW MDB but electrofishing is considered the most effective in sampling a full range of species and size classes in lowland rivers (Harris and Gehrke 1997). The inclusion of sites sampled with boat, backpack or both electrofishing methods (‘hybrid’ sites) sites ensures a range of habitats (main channel, riffles etc.) were sampled within the rivers.

To generate a dataset suitable for analysis, all boat and backpack electrofishing data from within the NSW MDB were extracted within the time range of 1st July 1994–30th June 2023 (data beyond these bounds were too sparse to include). No samples were collected between 1st July 1996 and 30th June 1997. Data were filtered to only include records from surveys of fish community composition (excluding projects targeting specific species) from elevations < 700 m that caught at least one fish. Although carp occur above this altitude, ‘montane’ altitudes > 700 m are not considered key habitat for carp (See Driver et al. 1997).

Each record contained the catch for each species in an electrofishing survey, with electrofishing shots aggregated at the site level. This resulted in 3,100 sampling events across 1,144 sites (Table 1, Fig. 1). To convert the abundances of fish into biomass within each sample, we used length–weight relationships to convert fork length measurements to weight estimates. While length–weight relationships are known to vary temporally, in some years weight data was not collected and an initial analysis of carp revealed minimal differences, mainly in small fish (Figure S1). The length–weight relationships are therefore pooled across both time and space for each species and used all available data in the NSW DPI FER Database. For species count records where some individuals were sub-sampled and length data were not collected for every individual, the mean weight (converted from length) of the measured fish and number of each species in each sampling event was used to calculate biomass. To calculate the proportion of total fish biomass contributed by carp, we divided the carp biomass by the total biomass of all fish in the sample. Sample dates were assigned to sampling years based upon years starting in July and ending in June; for example, 2020 refers to the sampling year starting 1st July 2019 and ending 30th June 2020 (winter–winter). To generate datasets suitable for analyses limited to key rivers, the above data filtering steps were undertaken, except we selected only samples in main channel sites of nine key rivers (excluding tributaries and off-channel habitat), regardless of site altitude, as the objective was to derive river-scale estimates of carp biomass proportions (Table 1).

Table 1 Summary of the datasets used in each analysis
Full size table
Fig. 1
figure 1

Distribution of surveys used in the analyses of carp biomass as a proportion of total fish biomass. Bubble size shows the number of electrofishing surveys at a site while colour identifies the major rivers (white being samples that were used in the overall analysis but not a specific river analysis). The Murray–Darling Basin is shown in light grey

Full size image

Data analysis and visualisation

To visualise the distribution of sampling events (surveys) with various proportions of carp biomasses over time, we generated annual histograms from the NSW MDB dataset showing the proportion of sampling events with different proportions of carp biomass. We also calculated the proportion of surveys with < 10% and > 90% carp biomass each sampling year. Changes in these proportions over time were then tested with Bayesian generalised linear models using a beta error distribution to account for the proportional nature of the data. A change over time was considered significant if the estimate and 95% credible intervals (CI) did not overlap zero.

Bayesian generalised linear mixed models were used to quantify the proportion of fish biomass contributed by carp while accounting for unequal sampling effort over both time and space. To calculate our response variable for modelling, all replicate electrofishing shots within a survey were summed before calculating the proportion of total fish biomass comprised by carp. As the response variable (proportion of carp by biomass in each electrofishing survey) included both zeros and ones, we used a zero–one augmented beta error distribution following the guidelines of Douma and Weedon (2019). For the overall NSW MDB model, we included four random intercept effects: ‘Water Resource Planning Area’, ‘Sample Date’, Method and ‘Site’ and a fixed effect of ‘Sampling Year’. The ‘Water Resource Planning Area’ (WRPA) effect accounts for the nesting of sites within larger geographical regions while the ‘Sample Date’ and ‘Site’ effects account for dependency in the sampling design with multiple samples on the same day and repeated sampling at sites over time. The WRPA regions are management regions based on those defined by the Murray–Darling Basin Authority (Murray–Darling Basin Authority 2021), but modified slightly to match those used in Crook et al. (2023; Figure S2). For each river we ran a separate model of the same structure except for the exclusion of the WRPA random effect. All models were fitted within R v4.3.1 (R Core Team 2022), using the ‘brms’ package (Bürkner 2017, 2018), with weakly informative priors of normal(0,1). Model convergence was assessed visually and using R-hat statistics. All models included 10,000 iterations per chain (half as a warm-up).

To generate estimates of the proportion of carp to all fish biomass in each ‘Sampling Year’ for each model, we generated prediction plots for each species based upon the expectation of the posterior predictive distribution, not including the random effects using the R package ‘tidybayes’ (Kay 2021). This results in a distribution of estimates showing the likely true values. We produced annual estimates for the overall NSW MDB and each key river. We also extracted overall estimates of carp proportions for each WRPA. All data processing was conducted using the ‘tidyverse’ package (Wickham et al. 2019).

Results

Across the NSW MDB, the survey-level histograms revealed that there have consistently been surveys that only recorded carp (≈100% carp biomass) and surveys that recorded no carp or < 10% (0–10% carp biomass, Fig. 2). Most of the histograms were skewed towards moderate or high proportions of carp biomass. The beta regression of annual proportions showed that the proportion of surveys with > 90% carp biomass decreased significantly from approximately 33% (95% CI: 27–40%) of all surveys in the mid-1990s to approximately 20% (95%CI: 16–26%) at the end of the time series (Fig. 3; estimate: − 0.02 (95% CI: − 0.04 to − 0.01)). Surveys with < 10% carp biomass on the other hand were more stable across the time showing no evidence of an overall change at approximately 22% of surveys each season (Fig. 3; estimate: − 0.02 (95% CI: − 0.04 + 0.01)).

Fig. 2
figure 2

Histograms showing the percentage of surveys across the Murray–Darling Basin with different levels of carp biomass each sampling year (Year ending June 30th, elevation < 700 m). n indicates the number of electrofishing surveys undertaken each sampling year

Full size image
Fig. 3
figure 3

Proportion of surveys over time with < 10% (orange) or > 90% (blue) fish biomass being comprised of carp based upon all electrofishing surveys across the NSW MDB (elevations < 700 m). Symbol size represents the number of surveys undertaken each year. Year represents a sampling year ending June 30th of the specified year and beginning July 1st the previous year. The solid lines represent a beta regression with the shaded area showing 95% credible interval. Note there was no data for 1997

Full size image

When accounting for the dependencies in the data structure with the mixed effect model, the median proportion of carp from all fish biomass across the NSW MDB was relatively stable across the time series, with an overall median of 57.0% (95% CI: 42.5–70.4%). Annual median estimates ranged from 49.3% (95%CI: 37.6–60.3%) in 2008 to 67.1% (95% CI: 56.7–74.1%) in 1999, although the 95% credible intervals on these estimates overlapped in all years (Fig. 4).

Fig. 4
figure 4

Annual estimates of the percentage of biomass contributed by carp using data from sites across the New South Wales Murray–Darling Basin (elevation < 700 m). The dots show the median estimates, thick grey bars show the 80% credible intervals (CI), and the thin grey bars show the 95% CIs. The shading shows the distribution of estimates

Full size image

At the regional WRPA scale, there were differences among the areas. The Murrumbidgee had the highest estimated median proportion of biomass of carp, followed closely by the Lachlan and Macquarie-Castlereagh WRPAs, although the 95% CI intervals overlapped for all areas. The lowest median proportion of carp biomass was estimated to be in the Barwon-Darling Watercourse and NSW Border Rivers (Fig. 5).

Fig. 5
figure 5

Estimates of the percentage of biomass contributed by carp using data from sites within each of the NSW Water Resource Planning Areas. The dots show the median estimates, thick grey bars show the 80% credible intervals (CI), and the thin grey bars show the 95% CIs. The shading shows the distribution of estimates

Full size image

At the individual river scale, the Macquarie, Lachlan and Namoi Rivers consistently had the highest median proportion of carp biomass. The Gwydir, Darling-Baaka and Castlereagh Rivers had the lowest proportion of biomass contributed by carp (Fig. 6). The Darling-Baaka had very few surveys with > 90% carp and the Gwydir had a high proportion of surveys with < 10% carp biomass detected (Figure S3). In some years and rivers there was large uncertainty around the model estimates (e.g., Darling-Baaka, Murray and Murrumbidgee Rivers; Fig. 6) which can be attributed to low sampling effort in those years combined with highly variable data between surveys. Noting these large credible interval, median estimates fluctuated between years ranging from 3% in 1996 (where sampling effort was low and the estimate is considered unreliable) in the Murrumbidgee to 90.6% in the Macquarie River in 2007. Most rivers had relatively stable time-series of estimates although the Barwon, Lachlan, Murray and Gwydir rivers showed considerable fluctuations but no clear trend over the time series.

Fig. 6
figure 6

Annual estimates of the percentage of biomass contributed by carp using data from sites on key NSW rivers. The dots show the median estimates, thick grey bars show the 80% credible intervals (CI), and the thin grey bars show the 95% CIs. Estimates were made using predictions from the fitted models ignoring the random effects and assuming the more common boat electrofishing method

Full size image

Discussion

The results of the current study demonstrate that over the last three decades, carp have comprised a median of ≈ 57% of fish biomass at sites across the NSW portion of the MDB (at elevations < 700 m) and this proportion has been stable over the past three decades. It is of great concern that an invasive species has occupied such a large proportion of biomass in the ecosystem over the past three decades, and there is no doubt that the impacts of carp have contributed to declines in the ecological condition of rivers across the MDB over that period (Koehn 2004; Marshall et al. 2019). Our results show that carp biomass currently exceeds 90% in  ≈ 20% of electrofishing surveys, with no evidence of an increasing trend in such sites. A similar proportion of sites had less than 10% carp biomass. The stability of carp as a proportion of total fish biomass over the time series at a large spatial scale, combined with recent research showing fluctuations with no clear trend in carp abundance and biomass over recent decades (Crook et al. 2023), support the hypothesis of logistic population growth. This suggests that carp in this region may have been operating at their ecosystem carrying capacity since at least the 1990s following their rapid spread in the 1970s and 1980s (Forsyth et al. 2013).

Comparison with previous estimates

Our results largely align with previous work, however there are some differences in our estimates that can be attributed to methodology. Two relevant pieces of precious work are the large scale Sustainable Rivers Audit (SRA) surveys undertaken between 2007 and 2010 (Davies et al. 2012a, b), and the detailed catchment specific monitoring of fish communities by NSW DPI (e.g. Gilligan 2005). We focus our comparisons on these studies representing basin-scale and catchment-scale estimates but note many studies have made similar estimates. The comprehensive SRA surveys found that alien species contributed an average of 61% of total fish biomass, with most of this carp (Davies et al. 2012b, a). The SRA program had a spatially stratified sampling design, with multiple sampling methods (electrofishing and traps), employed to represent different habitats (including a quarter of sites being > 700 m where carp biomass is generally lower) at valley and Basin scales so that large scale conclusions could be drawn with relatively simple analyses. In contrast, the current analysis focused only on electrofishing and included many more samples and sites. Gilligan (2005) estimated that carp contributed 87% of total fish biomass based on data summed across 28 sites sampled in the Murrumbidgee catchment in 2004 (total observed carp biomass divided by total biomass of all observed fish). This estimate is higher than both our estimate for the overall Murrumbidgee WRPA (66%) and the 2004 estimate in the Murrumbidgee River (52%). The sites included in Gilligan (2005) were randomly selected following the same process as the SRA surveys but the calculations did include sites with very high carp biomass from the Tumut River in the Australian Capital Territory which the current study (focussed on NSW) did not include and the inclusion of such sites would likely increase the overall carp proportion.

The analytical treatment of fish survey data has the potential to considerably influence estimates of the contribution of carp to total fish biomass and conclusions drawn (Gould et al. 2023). In addition to the analyses presented, we conducted analyses where shots were modelled separately (nested within sampling events) rather than being summed within each site. These analyses produced substantially lower estimates of median proportion carp biomass (≈ 40% at the Basin scale). This difference in analytical outcome was due to the extreme heterogeneity of biomass (both total and carp) between shots within a site (< 100 g–80 kg), and the general tendency for only a small minority of shots to have very high carp biomass. When proportions were calculated using shot level data, substantially more shots with low carp biomass contributed to the estimates, resulting in lower median values. While analysis of shot level data is legitimate depending on the research question, we consider summing at the site level the most intuitive approach to model the proportion of biomass comprised by carp. Nonetheless, the difference in outcomes resulting from slightly different analytical approaches demonstrates the importance of detailed descriptions of calculation methods in studies of this type (Gould et al. 2023).

It should be noted that electrofishing efficiency varies by species and therefore some species may be consistently over or under-represented in our sampling. For example, Murray cod and Golden perch have higher capture probabilities using electrofishing than Silver perch (0.24 and 0.08 capture efficiency compared to 0.005; Lyon et al. 2014). The capture probability for carp is likely to be similar to golden perch but less than Murray cod (≈ 0.05–0.1) based upon North American studies (Bayley and Austen 2002), although it can vary substantially across habitat types (Stuart et al. 2021). The use of generalised linear mixed models in the current study controlled for unequal sampling effort, repeated sampling at some sites and varying sampling efficiency between sites and days. This allowed for a larger dataset (both spatially and temporally) to be analysed and reliable estimates of the mean proportion of biomass comprised by carp in the NSW portion of the MDB to be produced. While it is possible that the models systematically underestimate or overestimate the dominance of carp, we consider it unlikely that our analyses are strongly biased because carp have a capture probability in the range of most of the large-bodied fish included in the estimates.

Some previous estimates of carp biomass have considered wetland habitats as they are key habitat for carp breeding and recruitment (Stuart and Jones 2006; Rayner et al. 2015; Stuart et al. 2021). A limitation of the current study is that our dataset under-represents wetland environments due to difficulty in sampling many of these habitats with a comparable method to river channels, leading to them being excluded from many projects in our database. Further investigation is warranted regarding the species composition of wetland habitats where it is possible that the proportion of fish biomass contributed by carp may be higher (Stuart et al. 2021). Our analysis also targeted elevations carp are known to prefer (< 700 m). It is likely that the median proportion of carp biomass would be lower if all elevations were included because carp are rarely found in high abundance in high elevation streams.

It should also be noted that our study focused on the NSW portion of the MDB. Previous large-scale surveys such as SRA surveyed the whole MDB and reported catchment level results for percentage of biomass comprised by alien species. In the SRA surveys, large variation was reported between catchments (32–87%). While some of our estimates for key NSW rivers align closely with reported alien biomass proportions from the SRA (Gwydir, NSW Border Rivers), our estimates for some regions were up to 12% different (Barwon-Darling Watercourse, Murrumbidgee) (Davies et al. 2012b, a). This may be due to a variety of factors including the SRA including other alien species. As our estimates largely align with the median values in the raw histograms, we are confident that our analyses provide reasonable estimates of the median contribution of carp to total fish biomass in these rivers, noting there are sites across the whole spectrum of carp dominance (from 0–100% biomass).

Ecological and management implications

In Australia, carp is a high-priority pest species and the Australian government is considering management options in parallel with additional research identified in the National Carp Control Plan (NCCP 2022). Our estimates of the contribution of carp to total fish biomass can facilitate informed debate surrounding management of carp in the MDB. Based on our analysis, it appears that carp comprise 43–70% of fish biomass across the NSW MDB (with median values varying from ≈ 45–66% at individual valley scales and ≈ 46–80% at individual river scales). The river scale analysis in this study demonstrates the importance of smaller scale spatial estimates which may be more relevant for management. This heterogeneity is highlighted by the Darling-Baaka and Barwon rivers which have vastly differing estimates yet are both within the broader Barwon-Darling Watercourse WRPA management area. The Gwydir River and Gwydir WRPA represents another interesting case study with the Gwydir River above Copeton Dam being one of the largest known areas within the MDB where carp have not yet become established. This area likely drove the relatively low estimate of average proportion of carp biomass in this river. There also appears to be a nonlinear pattern in the Gwydir River within a possible decline in the late 2000s before an increase back to the 2000 level in late 2010s. The maintenance of carp-free areas such as the upper Gwydir catchment should be viewed as a high management priority (Beatty et al. 2017).

While there was an initial ‘explosion’ of the carp population after the introduction of the ‘Boolarra’ strain (Forsyth et al. 2013), major fluctuations in abundance over the time series appear to correspond to large recruitment events (usually corresponding with floods). While these events are associated with dramatic increases in carp abundance (Rayner et al. 2015; Stuart et al. 2021), changes are driven by high numbers of small recruits and carp abundance has generally returned to pre-flood levels within 2–3 years, presumably due to high juvenile mortality and ecosystem carrying capacity constraints (Pearson et al. 2022; Crook et al. 2023). This is likely similar to the sporadic ‘booms’ in populations of Gambusia affinis in central USA, that are thought to be triggered by anomalous warm spring temperatures which drive early and successful reproduction in some years, with the population returning to pre- ‘boom’ levels in the following years as temperatures normalise (Matthews and Marsh-Matthews 2011).

Carp are generally classified as a ‘periodic’ species (Winemiller and Rose 1992; Olden et al. 2006), and exhibit strong juvenile recruitment in response to high flow events that inundate riverine wetland habitats used for spawning (Stuart and Jones 2006). It was somewhat surprising, therefore, that the median proportion of total biomass contributed by carp across the NSW MDB was relatively stable across the time series we analysed. This is despite a series of flood years that resulted in high levels of carp recruitment (e.g., 2010–2012, Crook et al. 2023) and a temporary increase in sites dominated by carp. It appears that flood-driven booms in carp recruitment do not persist as long-term changes in proportional biomass, likely due to native fish biomass also responding positively to floods and offsetting the increase in carp as a proportion of overall fish biomass. This suggestion is supported by extensive previous research reporting positive associations between high river flows and increased native fish recruitment and growth (e.g., Humphries et al. 1999; Balcombe et al. 2009; Zampatti and Leigh 2013; Stoffels et al. 2020; Tonkin et al. 2021). The stability of total fish biomass contributed by carp may also suggest that they do not fit neatly within the periodic life history category but, instead, may exhibit characteristics of both periodic and equilibrium species. Although there was no evidence of a temporal trend at the Basin scale, the results showed temporal fluctuations in carp biomass proportions in some rivers. Most notably the Barwon River appeared to show a recent increase in carp dominance whereas the Murray exhibited a recent decline. Other rivers generally had median annual estimates that fluctuated within ~ 25%. It may be that while carp overall in the NSW MDB are operating at their carrying capacity, populations are fluctuating in rivers, potentially due to temporal fluctuations in environmental conditions, demonstrating the local carrying capacity is variable dependent on the conditions.

The current analysis, along with the time series of abundance and biomass presented in Crook et al., (2023), extends the population growth phase shown by Forsyth et al., (2013). Fluctuations with no clear trajectory over recent decades in carp abundance and biomass (as shown in Crook et al. 2023), and the stability of the median proportion of total fish biomass comprised by carp (Fig. 4), provide multiple lines of evidence to suggest that the biomass of the NSW carp population is limited by carrying capacity constraints. This aligns with previous predictions that the population may have been following a logistic population growth curve based upon data from the 1960s to the early 2000s and an apparent fluctuating population growth rate over time (Forsyth et al. 2013). It is likely that given enough time following introduction, most invasive species will reach a carrying capacity and stabilise. For example, lionfish (Pterois volitans) in the Gulf of Mexico started to display density-dependent impacts less than seven years after first appearing, suggesting that the population rapidly reached carrying capacity (Dahl et al. 2019). Our findings for carp in the NSW MDB have similarities to the lionfish example, with community composition appearing to have stabilised as the species has become fully established. It would be interesting for future MDB studies to quantitatively analyse community structure, possibly using joint-species distribution models, ecosystem models or other multivariate analyses to better understand inter-species relationships.

Whilst challenging, there are examples of stability in community composition being disrupted by dramatic changes in the population or environment. In the upper Mississippi River, USA, a long-established carp population began to decline dramatically in the 1980s, while at the same time native bluegill increased in abundance (Gibson-Reinemer et al. 2017). It is believed that this near-spontaneous collapse of the carp population and shift in the fish assemblage was driven by an unplanned introduction of a carp herpes virus CyHV-1 or CyHV-3 (Gibson-Reinemer et al. 2017). The proposed release of the CyHV-3 virus in Australia to control carp has attracted much recent debate, with strong advocacy by scientists both for (e.g., McColl et al. 2014, 2017; McColl and Sunarto 2020; Stuart et al. 2023) and against its release (e.g., Lighten and van Oosterhout 2017; Kopf et al. 2017, 2019a; Marshall et al. 2018), and considerable public interest in this issue (Mehmet et al. 2018). If an introduction of the CyHV-3 virus was to be implemented in the MDB, it is likely that considerable changes in fish assemblages would occur over large spatial scales, with important consequences for food webs and broad-scale ecosystem processes. Our estimates of the relative contribution of carp to total biomass have the potential to inform analysis and debate on this important issue.

The recent estimates of carp biomass by Stuart et al. (2021) suggest that during an average hydrological scenario there is 42,936t (95% CI: 23,055–77,769) of carp biomass in eastern Australian rivers (mostly in the MDB; this estimate excludes wetlands, lakes, storages etc.). Estimates of actual biomass such as that by Stuart et al. (2021) are important to put proportional analyses such as the current study into context but need to be made with caution as all sampling methods including electrofishing suffer from biases such as varying detectability (Lyon et al. 2014). To accurately make estimates of biomass (either absolute or relative), these need to be controlled for, either through explicitly modelling the varying detectability (such as in Stuart et al. 2021), or through techniques such as the random effects used in the current study. The current study did not attempt to model absolute biomass. Moving forward analyses of both proportional and actual biomass of other species will be important to provide ecosystem level understanding of waterways across the MBD, particularly as these estimates can vary substantially at spatial scales from within sites to across the whole MDB. Understanding the biomass of other species would also enable the creation of ecosystem models, potentially enabling larger scale ecosystem management (Geary et al. 2020).

Conclusions

While there are sites across the NSW MDB where > 90% of fish biomass is comprised by carp, our results demonstrate that reference to this statistic without additional context is potentially misleading. The prevailing emphasis on the “up to 90% carp biomass” statistic in the media and scientific literature focusses on the upper extreme of reported data. A more accurate representation of the data would emphasise that a median of 57% of fish biomass is comprised by carp across samples taken in low-elevation sites in the NSW MDB, and that this proportion has been stable over the past three decades.

In addition to providing quantitative information to inform a balanced debate on the relative merits of the available management options for carp, our estimates present an opportunity for engagement with local communities regarding the current state of fish communities (Mooney and Tan 2012). While  ≈ 20% of sites have > 90% biomass carp, the fact that  ≈ 20% of sites have < 10% carp biomass demonstrates that there are still significant areas of the NSW MDB where native fish populations remain in reasonable health (see also Crook et al. 2023) and provides justification for increased investment in the protection of these ecologically valuable populations and habitats.

At the time of writing, juvenile carp numbers across the MDB were high following recent floods from 2020–22, that no doubt further fueled perceptions of carp dominance across the landscape. However, based on analyses of available data since the 1990s (Crook et al. 2023), carp numbers are likely to decline once more (before any substantial increase in adult carp biomass) as the climate moves into a drier climatic phase. While temporary increases in invasive species abundance often elicit calls for urgent action, the example of carp in the MDB demonstrates the importance of considering temporal dynamism in invasive species populations in a long-term context, thus facilitating development of management responses that match the time frames over which the populations operate.