Contemporary Trends in the Spatial Extent of Common Riverine Fish Species in Australia’s Murray–Darling Basin

Abstract

As one of the world’s most regulated river basins, the semi-arid Murray–Darling Basin (MDB) in south-eastern Australia is considered at high ecological risk, with substantial declines in native fish populations already identified and climate change threats looming. This places great importance on the collection and use of data to document population trends over large spatial extents, inform management decisions, and provide baselines from which change can be measured. We used two medium-term data sets (10 MDB basin-wide fish surveys from 2004–2022) covering the 23 catchments and 68 sub-catchments of the MDB to investigate trends in the distribution of common riverine species at the entire basin scale. Fifteen native species were analysed for changes in their contemporary range, and whilst short-term changes were identified, all species showed no significant continuous trend over the study period. We further analysed the native species extent relative to their historic records, with bony herring and golden perch occurring in 78% and 68% of their historic river kilometres, respectively, whereas southern pygmy perch, northern river blackfish, silver perch, mountain galaxias, and freshwater catfish were all estimated to occur in less than 10% of their historic extent. Six established non-native species were also analysed and were very consistent in extent over the years, suggesting that they are near the available limits of expansion of their invasion. We provide effect sizes for the spatial extent index which can be used as baselines for future studies, especially those aiming to monitor changes in the spatial extent and population status of native species, or changes in the spatial extent of new or existing non-native species.

1. Introduction

The Murray–Darling Basin (MDB) is one of the world’s most regulated river basins [1,2], with the majority of its 23 catchments classified as having impaired connectivity [1] and considered in poor ecological condition [3,4,5]. Over-allocation of water, flow regulation, and environmental damage have been long-term concerns, e.g., [6,7,8], which are becoming more pressing as a result of climate change, e.g., [9,10,11]. Recent severe drought conditions, extensive fish kills, and extreme bushfires (2019–2020) have heightened concerns over the ecological health of the MDB [12,13,14]. Even prior to these recent concerns, MDB native fish were known to be in serious declines, with populations (viz distribution and abundance) of common riverine species estimated to be at <10% of pre-European settlement levels (early 1800s) [15,16].

Recognition of the need for higher confidence in native fish species status together with considerable investment in initiatives to improve river health, e.g., [17,18], require a significant monitoring program. Hence, the MDB Sustainable Rivers Audit (SRA) began in 2004 as an MDB-wide surveillance monitoring program to report on the status and long-term trends in the ecological health (condition) of rivers at the river valley (catchment/watershed) scale. One of the five ecological health themes included in the SRA was fish and this included the first ever attempt at an MDB-wide fishery-independent assessment program [4,5]. The SRA fish sampling methodology aimed to assess riverine species richness, only sampled in channel-permanent habitats (no floodplain or ephemeral habitats), and relied on a single sampling technique (electrofishing) [4,19]. Nine years after the SRA program was implemented, the fish component was transformed into the MDB Fish Survey (MDBFS) which revisited sites previously used in the SRA and maintained the same within-site sampling protocols but sampled fewer sites more often. Both data sets combined provide the first medium-term (10 basin-wide surveys over 19 years; 2004–2022) data set for fish in the MDB. The first 9 years of data have previously been used to report on river health using conceptual indicators of health, e.g., [5], and response to hydrological factors [19], but these data have not been widely analysed as a combined data set nor used to investigate the population trends for MDB fish species over time (see [20]).

Monitoring fish distributions is a useful measure of how populations are responding to management interventions (e.g., fish passage, habitat rehabilitation) or how climatic factors (drought, flood) are influencing populations. It is also a component of assessing conservation status and a key metric in assessing the recovery of threatened species where extent of occurrence and area of occurrence are widely used in classifications such as the IUCN red list [21]. The MDB also has a high proportion of non-native fish species, both from overseas and translocated native species [22]. Although several non-native species have their distribution in the MDB limited by environmental variables (e.g., Salmonids and water temperature), others (including those recently introduced) are habitat generalists (e.g., common carp Cyprinus carpio) with considerable potential for range expansions [22]. Non-native fish are significant hosts and vectors of novel pathogens [23,24], and tracking the extent of the distribution of non-native fish is a key concern for managers. Non-native species are an ongoing threat that may have high social capital but are rarely targeted in river health monitoring programs [25].

Aims of this Study

We used the SRA and MDBFS data sets to determine the trends in the contemporary spatial distribution (extent) of common, widespread, and abundant MDB fish species over the period 2004–2022. We further assessed spatial extent for each native species relative to their known historical distribution and the current spatial extent of non-native (Non-native) fish species in the MDB. The results provide trends and baseline assessments of the extent of common riverine species that may be used as reference for future comparisons.

2. Materials and Methods

2.1. Study Area

The Murray–Darling Basin in south-eastern Australia (Figure 1) covers more than a million km², or about 14% of Australia’s total landmass, provides about 40% of Australia’s total agricultural production, and accounts for 50% of the nation’s irrigated agricultural water use [26,27]. Hence, there is conflict between water management for agriculture and that for environmental benefits and assets such as fish [28]. The MDB encompasses all or parts of five states and territories, with fish management remaining the responsibility of state/territory governments. Overlaying this structure is a Commonwealth responsibility for water management via the Basin Plan [27]. The MDB is subdivided into 23 valleys/catchments/watersheds and 68 sub-catchments for water use management and for SRA/MDFS fish sampling (Figure 1).

2.2. Data Sets

The MDBFS and SRA data sets offer medium-term data covering the whole of the MBD stream network with a standardised sampling effort for fish based on a probabilistic site selection strategy in a mixed longitudinal rotating panel design [29]. The sampling frame covers all streams in the MDB with at least 5 GL mean annual flow and does not include ephemeral or non-riverine (wetland and floodplain) habitats. Each valley is stratified into three sub-catchments that align with potential changes in fish assemblages. These sub-catchments include lowland (<200 m ASL), slope (200 < 600 m ASL), and upland (>600 m ASL) bioregions. Upland sites are sampled with relatively more spatial intensity than lowland or slope habitats [19]. Consequently, when aggregating to higher scale assessments (catchment or MDB), site parameters must be weighted by the selection probability of each site (weighting is easily interpreted as the kilometres of river represented by the site when it first enters the sampling program). Sites are 1 km in length and were selected from a stream length GIS layer; thus, after aggregating site data to larger extents, the assessments are best interpreted as river kilometres (e.g., fish/km) rather than as averages.

The SRA sampled more intensely, but less frequently, whilst the MBDFS sampled more frequently but less intensely:

• SRA pilot study (2003) = four MDB valleys and sub-catchments were sampled once with 21 to 26 sites per valley. A total of 92 sites sampled; the results are summarised in [19,30];
• SRA (2004 to 2013) = every MDB valley and its sub-catchment was sampled once every three years with 14 to 28 sites per valley. Approximately 450 sites sampled per year;
• MDBFS (2014 to 2022) = every MDB valley and its sub-catchment was sampled once every year except 2019 and 2020, during which half of the MDB was sampled. Approximately 145 sites were sampled per year (except 2019 and 2020), with 4–8 sites per valley, and all sites were previously sampled SRA sites.

Combined, the data sets include 1222 separate survey sites (each site represents 1 km of river) and 2368 sampling events. We treat sampling the whole MDB in single or consecutive years as a sampling round and, after 10 sampling rounds, the combined data set now offers the opportunity to analyse large spatial extent trends in the frequency of occurrence (extent) for common species that are ubiquitous and/or abundant in the data set.

2.3. Spatial Extent of Common Fish in the MDB (2004 to 2022)

We choose spatial extent as a practical measure for this data set because whilst abundance is also important, its estimation by any sampling method is difficult and comparing abundance through time is susceptible to many confounding factors. For example, fish have varying susceptibility between species (habits, habitats, and sizes) and within species (e.g., life phases) to different sampling seasons and methods, and detectability varies with environmental variables such as flow and turbidity. This makes the assessment of abundance best interpreted as a relative assessment, e.g., [31,32,33]. The extent of occurrence, on the other hand, the proportion of river sites or river kilometres where a fish occurs, offers a simpler assessment of fish population health over a large area as it only requires detecting the species presence as a standard effort. It relies simply on the assumption that the detection of the presence of a species within a site is more likely when the species is more abundant. This assumption is clearly the case for many fish species in the MDB when sampled via electrofishing (refer to Table 2 in [31]).

The within-site sampling methodology of the SRA was designed to target species presence rather than to measure abundance. It involved intensive electrofishing and returned the full list of species that are well-sampled by electrofishing in 174 of 180 (97%) of sites in the first year of implementation [19]. Thus, the data collected by the SRA/MDBFS programs are highly suitable for estimating the presence of common main-channel species that are susceptible to electrofishing within each site [19] and, consequently, their extent at the sub-catchment and basin scale. This includes most large-bodied species and riverine species that have a wide distribution in the MDB.

There are three aspects to the spatial extent analyses. Firstly, for each native species, we identified contemporary sub-catchments as sub-catchments where a species had been collected at least once since 2004 were also a historically (pre-1980) known sub-catchment for that species, based on [22]. As we use sites where the species had been collected at least once in the data sets, we treat this analysis as a prevalence assessment and refer to the results as an assessment of current or contemporary extent. We ignore sub-catchments where there was no history of occurrence to avoid misinterpretations from post-1980 translocations. In a second analysis, we also include an estimated absolute extent for each species relative to the species’ known distribution [22]. In this calculation, sites that are in a zone where the species should be present are included, regardless of whether the species has ever been collected in that particular site. This is considered a conservative estimate as the sites are randomly selected, and a missing species may be a random effect (e.g., that site has an incorrect habitat for that species). Nevertheless, it is an empirical estimate that returns a relative assessment of the species’ overall status that serves two purposes: (1) it allows us to identify species that have relatively better or worse spatial extent estimates than other species, and (2) it serves as a baseline for these species for this study period to allow managers to make future comparisons. The two assessments should be interpreted in conjunction. For example, a species may appear to have high contemporary prevalence—by consistently occurring in a number of sites in the current data set—yet those sites may only be from a fraction of its historic distribution.

On the other hand, we analyse non-native species relative to the entire sampling frame. In other words, the trend in extent of each non-native species is analysed relative to their distribution across the entire MDB, but for native species, trends in extent are relative to their historical distribution. In summary, the non-native species analyses include all 68 sub-catchments, whereas the number of sub-catchments used for native species are unique to each species.

To avoid potential confounding between the SRA and MDBFS site composition, only sites from the SRA program that were also sampled at least once in the MDBFS are included in the analyses. That is, many of the SRA sites were only sampled once, whilst almost all MDBFS sites were revisited annually or biennially, and all MDBFS sites were also sampled in the SRA. Single-visit SRA sites are not included in the analyses because they only occurred in the first three sampling rounds and could not contribute to a long-term trend analysis. For example, a single-visit site that that did (or did not) have a particular species collected can never contribute to the trend analysis because it is never sampled again. Meanwhile, a repeat-visit site that did (or did not) have the species collected can have the species collected (or not) in the repeat visits and can therefore contribute to the trend analysis of change in extent. There are 169 sites that were sampled in both the SRA and the MDBFS and these offer 1222 sampling events that are used in the current paper. The data do not include the pilot SRA as the sampling frame and site selection strategy changed after the pilot.

Only species that had been detected in at least 14 of the 68 (20%) MDB defined sub-catchments are included in the analyses. The 20% occurrence requirement returns enough data points to reasonably estimate the extent of each sampling round with 95% confidence intervals and constrains the analyses to species well collected by electrofishing. If such a species is abundant within a site, it is expected to be collected by the electrofishing survey. Accordingly, if a common species was not collected in a site or sub-catchment, it was either absent or in low abundance.

The data are assessed at the entire basin scale over 10 sampling rounds. The first 9 years (SRA) of data represent 3 rounds where the entire basin was sampled, whilst the remaining 8 years (MDBFS) represent 7 entire basin sampling rounds. That is, the first 3 rounds cover 3 years each and the remaining rounds are 1 or 2-year periods.

2.4. Statistical Analyses

The spatial extent index is the weighted average proportion of river kilometres in which each species was detected during each sampling round and is described in Equation (1).

$$I_{{extent}_{i,r}}=\frac{\sum d_{i_{\left[y=1\right]},r}}{\sum d_{{{\mathrm{E}}_i}_{\left[y=1\right]},r}}$$

(1)

where i = species, r = sampling round, d_i = river kilometres sampled for species i, y = 1 is where the species was detected, and dE_i(y=1) is the sampled river kilometres where species i was expected to be detected.

For native species, the denominator of Equation (1) is different for the prevalence analyses (expected river kilometres are constrained to known contemporary river kilometres) and the historical analyses (expected river kilometres include pre-1980 river kilometres). For non-native species, the denominator is set as the entire basin river kilometres sampled each round.

To estimate trends, the index was estimated for all rounds using a generalised linear model fitted with a binomial response (species presence = 1, absence = 0) where sites are random replicates within sub-catchments which are repeated subjects with an independent correlation structure. Because there are differences in sampling efficiency between species, the index should be interpreted in a relative, not definitive, manner when comparing between species.

When graphing trends, the calculated index of extent for prevalence (contemporary distribution) for native species and overall extent for non-native species was plotted as a time series with a modelled three-period centred moving average for the index and its 95% confidence intervals. We assessed monotonic trends in the spatial index for each species over the 10 sampling rounds using the Kendall rank correlation test. Complex trends were not tested statistically as there were only 10 sampling rounds, and intervals are not strictly equidistant given each data points represent a 1, 2, or 3-year cycle. To investigate short or long-term changes in extent, we performed pairwise comparisons of spatial extent between every survey round for each species. To moderate the type-1 error rate (n = 45 comparisons per species), we only reported survey rounds where the comparison is significant at p < 0.01. To provide a baseline guide for monitoring, we estimated the effect size for the spatial extent index between survey rounds for each species. We plotted the magnitude of differences in extent between all sampling rounds with the probability of each difference being significant and fit a LOESS (locally estimated scatterplot smoothing) line to the plot. We estimated the effect size as the magnitude of difference between sampling rounds where the LOESS intersected with the 0.05 level of significance. We have provided a guide to the sensitivity of the spatial index for monitoring, calculated as the effect size expressed as a percentage relative to the long-term average. Trends for the historical analyses for native species are included on the same plots to aid in interpreting status comparison, and we have provided a summary table of all species extent relative to their historical distribution. All analyses were performed using SAS/STAT^®14.1 [34].

3. Results

3.1. Common Fish Trends in Extent in the MDB (2004 to 2022)

Twenty-two fish species were detected in at least 20% of the MDB sub-catchments and 21 species were analysed for trends in spatial extent (

Table 1. Common fish species in the MDBFS/SRA data set between 2004 and 2022. Several Hypseleotris species were combined for the analyses because of taxonomic resolution differences in the early years of the monitoring programs, e.g., [35,36]. * Obscure galaxias: Galaxias oliros were not analysed for trends in spatial extent. The number of sub-catchments collected are cumulative over the 10 sampling rounds and used for the prevalence analysis. The number of historical sub-catchments expected are based on [22] and used to estimate overall status compared to historical status for native species. The effect size of the index is the magnitude of change in contemporary extent between any two sampling rounds that would be considered statistically significant (p < 0.05). The % change in the index is the effect size relative to the 20-year average from the current study. na = non-estimable.

Species	Common Name	Origin	Life Guild	Num. Sub-Catchments Collected	Num. Sub-Catchments Expected	Index Effect Size	Index % Change
Gambusia holbrooki	eastern gambusia	Non-native	Short-lived	60	0	0.17	36%
Cyprinus carpio	common carp	Non-native	Long-lived	54	0	0.09	11%
Retropinna semoni	Australian smelt	Native	Short-lived	51	66	0.23	48%
Carassius auratus	goldfish	Non-native	Long-lived	52	0	0.17	45%
Hypseleotris spp.	carp gudgeon complex	Native	Short-lived	47	56	0.17	31%
Macquaria ambigua	golden perch	Native	Long-lived	49	66	0.16	24%
Perca fluviatilis	redfin perch	Non-native	Long-lived	44	0	0.08	61%
Maccullochella peelii	Murray cod	Native	Long-lived	43	63	0.19	39%
Nematalosa erebi	bony herring	Native	Intermediate-lived	29	49	0.15	19%
Philypnodon grandiceps	flathead gudgeon	Native	Intermediate-lived	22	30	0.24	68%
Tandanus tandanus	freshwater catfish	Native	Long-lived	23	50	0.32	>100%
Galaxias olidus	mountain galaxias	Native	Intermediate-lived	26	51	0.4	89%
Salmo trutta	brown trout	Non-native	Long-lived	27	0	0.05	75%
Oncorhynchus mykiss	rainbow trout	Non-native	Intermediate-lived	26	0	0.05	94%
Gadopsis marmarata	northern river blackfish	Native	Intermediate-lived	20	53	0.18	28%
Melanotaenia fluviatilis	Murray–Darling rainbowfish	Native	Short-lived	23	46	0.26	46%
Leiopotherapon unicolor	spangled perch	Native	Intermediate-lived	23	33	0.28	53%
Bidyanus bidyanus	silver perch	Native	Long-lived	21	51	0.26	>100%
Craterocephalus stercusmuscarum fulvus	unspecked hardyhead	Native	Short-lived	20	40	na	na
Galaxias oliros *	obscure galaxias	Native	Intermediate-lived	21	41	na	na
Gadopsis bispinosa	two-spined blackfish	Native	Intermediate-lived	18	22	0.3	43%
Nannoperca australis	southern pygmy perch	Native	Short-lived	14	36	na	na

). These included 15 native and 6 non-native species. Obscure galaxias were omitted from the analysis as they were undescribed at the start of the monitoring program—hence, no historical records. Non-native and native species are reported separately, and native species are grouped by life expectancy guild (Table 1) (long, intermediate, and short-lived species) to aid in interpretation. Short-lived species have life cycles of <3 years, intermediate-lived species have lifecycles from 3 to 6 years, and long-lived species have life cycles of 6 years or greater.

No species showed statistically significant monotonic trends throughout the 10 sampling rounds (p > 0.05). Golden perch (Macquaria ambigua) tended to show a consistent increase in spatial extent throughout the study period (Kendall’s Tau = 0.47, p < 0.07), whereas two-spined blackfish (Gadopsis bispinosa) tended to decrease in extent (Tau = −0.42, p < 0.09). The effect size for assessing short-term changes in the spatial extent index varied considerably among species and tended to be smaller in species that were collected more often (Table 1). Ten of the twenty-one species had effect sizes less than 50% of the long-term mean, with bony herring (19%), golden perch (24%), and common carp (11%) being the most sensitive.

3.2. Short-Lived Native Fish Species

• Short-lived species all showed high inter-annual variability in their index of contemporary spatial extent (Figure 2).
• The carp gudgeon complex (Hypseleotris spp.) was estimated to occur in about 55% of its current distribution throughout the study, but the year-to-year variability in extent was high (range from 34% to 66%) (Figure 2a). The contemporary extent of carp gudgeons in 2010–2013 and 2020/21 was significantly lower compared to all sampling rounds from 2014 to 2020 (p < 0.01).
• Murray–Darling rainbowfish’s (Melanotaenia fluviatilis) contemporary spatial extent was also highly variable between rounds (39% to 75%), and no persistent trends were found. The largest contemporary extent for the species was 75% in 2016/17, which was significantly higher than the smallest extents of 2010–2013 and 2016/17 (p < 0.01) (Figure 2b).
• Southern pygmy perch (Nannoperca australis) decreased from a three-round average spatial occurrence of 68% at the start of the data set to an average of 53% over the final three rounds (Figure 2c), but the trend was not statistically significant (p > 0.05). The lack of significance and low effect size for this species can be partially attributed to wider confidence intervals because of the low number of sub-catchments that it was collected in (Table 1).
• Unspecked hardyhead (Craterocephalus stercusmuscarum fulvus) was generally less variable than the other short-lived species between years (46% to 66%) and averaged a contemporary extent of 55% for the study period. The effect size was non-estimable.
• Australian smelt’s (Retropinna semoni) contemporary spatial extent was between 37% and 52% during the first nine sampling rounds but increased to 72% in 2021/22, which was significantly (p < 0.01) greater than in 2007–2013, 2017/18, or 2019–21 (Figure 2e).

When considering the overall status estimates compared to historical records (lower dotted lines on Figure 2), the carp gudgeon complex (Figure 2a) had the highest spatial extent, 40–60%, of historical river kilometres compared to the other short-lived species throughout the study. Southern pygmy perch was regularly only collected in less than 5% of its historical river kilometres (Figure 2c).

3.3. Intermediate-Lived Native Fish Species

• Bony herring (Nematalosa erebi) had the highest contemporary spatial extent index of all native species and varied between 62 and 92% between sampling rounds (Figure 3a). The index achieved 92% in 2016/17 and 2021/22, which was significantly higher (p < 0.01) than the lower scores in 2017/18 to 2020/21 (Figure 3a).
• Flathead gudgeon (Philypnodon grandiceps) was very consistent in estimated contemporary spatial extent, averaging 35% throughout time (Figure 3b).
• Mountain galaxias averaged 45% and its peak extent of 78% in 2014/15 was significantly (p < 0.01) greater than that from 2010 to 2013 and in 2021/22 (Figure 3c).
• Northern river blackfish (Gadopsis marmarata) (average 65%) underwent slight but non-significant declines in their current extent in the last few years of the data (Figure 3d).
• Spangled perch (Leiopotherapon unicolor) was highly variable, occurring in between 21 and 80% of contemporary river kilometres in the study (Figure 3e). It had a significantly lower extent in 2004–2010 and 2018/19 than in 2019–2020 (p < 0.01) or 2010–2013 and 2021/2022 (p < 0.02).
• Two-spined blackfish (Gadopsis bispinosa) was also relatively consistent in contemporary extent, estimated to occur in ~70% of contemporary river kilometres throughout the study (Figure 3d). It also underwent slight but non-significant declines in its current extent in the last four rounds of the study (Figure 3d).

Compared to historic distributions, bony herring and spangled perch were the most widespread and were consistently collected in more than 60% and 40% of historical river kilometres, respectively (Figure 3). The other four intermediate-lived species were collected in less than 20% of their historical river kilometres, and notably, northern river blackfish and mountain galaxias were only collected in less than 5% of their historical river kilometres.

3.4. Long-Lived Native Fish Species

• Freshwater catfish (Tandanus tandanus) averaged 29% for the extent index throughout the study, with occasional dips to 17% and highs up to 51% (Figure 4a). The high variability between year-to-year estimates of spatial extent was high and the long-term average was low; consequently, the effect size for this species is more than 100% of the mean (Table 1).
• The contemporary spatial extent for golden perch averaged 68% and there was a visible but not significant overall increase in extent throughout the study (Figure 4b). The extent in 2016/17 (80%) and 2021/22 (84%) was significantly higher (p < 0.01) than in 2004–2007 (50%), 2007–2010 (61%), and 2017/18 (57%) (Figure 4b).
• Murray cod (Maccullochella peelii) averaged its contemporary distribution at 49% during the study (Figure 4c) and was significantly greater in extent (p < 0.01) from 2014 to 2017 than in 2010–2013 and 2018/19 (Figure 4c).
• Silver perch (Bidyanus bidyanus) was consistently low in river kilometres in which it was collected (average 19%) and had significantly low contemporary spatial extent from 2019 to 2021 (1%) compared to 2010–2013 (34%) and 2021/22 (32%) (Figure 4d).

In comparison with known historical distributions, golden perch was collected in more than 50% of its historical river kilometres in every round after 2010, and Murray cod was consistently collected in 40–50% of its river kilometres throughout the 19-year study period (Figure 4). On the other hand, freshwater catfish and silver perch were only ever collected in less than 10% of their historic distribution (Figure 4).

3.5. Non-Native Species

• All of the non-native species had consistent extent distribution throughout the study, as indicated by narrow confidence intervals and smooth trend lines (Figure 5).
• Common carp was the most collected non-native species in the data set and was estimated to occur in between 74% and 90% of river kilometres throughout the study period (Figure 5b). Carp showed a consistent but non-significant (p > 0.05) increasing trend in extent throughout the monitoring period. Nevertheless, the final sampling round in 2021/2022 (90%) was significantly higher (p < 0.01) than the sampling rounds in 2004–2007, 2007–2010, and 2014/15 (Figure 5b).
• Eastern gambusia (Gambusia holbrooki) was detected in 60 of the 68 sub-catchments (Table 1), but it was rarely detected in more than 50% of its river kilometres in any sampling round and always between 35% and 62% of total river kilometres (Figure 5c). It was significantly higher (p < 0.01) in 2015/16 and 2016/17 than from 2004 to 2013 and in 2018/19 (Figure 5c).
• Goldfish’s (Carassius auratus) spatial extent had several peaks and troughs between 25% and 51% and averaged at 38% of its river kilometres throughout all years (Figure 5d). Goldfish had a significant (p < 0.01) higher extent in 2007–2010 and 2016/17 than in 2014/15 and 2019–21 (Figure 5d).
• Redfin perch (Perca fluviatilis) (max 19%) and both brown and rainbow trout (Salmo trutta and Oncorhynchus mykiss) (<9%) generally occurred in low MDB river kilometres throughout the study period (Figure 5a,e,f). Redfin perch had a significantly lower spatial extent in 2014/15 than in the first and third sampling rounds (Figure 5f).

Overall, the non-native species appear stable in their current distributions, with only eastern gambusia and redfin perch showing some occasional inter-annual variability in spatial extent. Common carp is clearly the most widespread non-native species, followed by eastern gambusia and goldfish (Figure 5). Redfin perch and both trout species appear restricted in their ranges.

3.6. Species Summary of Baseline Assessments and Comparisons

As trends in all native species’ spatial extents generally did not display consistent change throughout time, we compared their relative estimated average annual spatial extents during the monitoring program (

Table 2. The average estimated spatial extent of 21 species in the Murray–Darling Basin between 2004 and 2022 using the SRA/MDBFS data set. The confidence in estimates reflects our perception of the sampling susceptibility of the species to electrofishing and whether it is predominantly a riverine species. The average extent for native species is the average percent of the historical (pre-1980) river kilometres that the species was collected in. For non-native species, the average extent is relative to the entire 64,000 km of stream kilometres in the sampling frame.

Species	Confidence in Estimate	Average Extent (%)	Interpretation of Extent
Native Species—relative to pre-1980 distribution
Short-lived
unspecked hardyhead	Medium	21	Substantially declined in riverine habitats.
carp gudgeon complex	Medium	48	Widespread and abundant.
Murray–Darling rainbowfish	High	33	Has declined and is now patchily distributed.
southern pygmy perch	Low	2	Rare in main channel riverine habitats when using electrofishing.
Australian smelt	Medium	36	Widespread and abundant in lowland habitats.
Intermediate-lived
two-spined blackfish	High	28	Declined and fragmented distribution.
northern river blackfish	Low	5	Declined significantly in larger streams but assessment confounded by historic taxonomy.
mountain galaxias	Medium	6	Greatly reduced, especially in lowland streams or where trout are present.
spangled perch	Medium	48	Widespread in Northern Basin but penetrate Southern Basin rarely.
bony herring	High	68	Widespread and abundant.
flathead gudgeon	Low	17	Poorly sampled in riverine habitats by using electrofishing.
Long-lived
Murray cod	High	35	Has declined but widely distributed. Stocked
golden perch	High	56	Remain widespread in lowlands. Declined in uplands. Stocked.
freshwater catfish	Medium	10	Substantially declined in riverine habitats.
silver perch	Medium	5	Substantially declined in riverine habitats. Stocked
Non-Native Species—relative to all MDB riverine habitats
Short-lived
eastern gambusia	Medium	48	Successful invader. Widely distributed.
Intermediate-lived
goldfish	High	38	Successful invader in lowland rivers and some slope regions.
rainbow trout	High	5	Distributed widely in cool upland streams. Stocked.
Long-lived
common carp	High	81	Highly successful and widely distributed.
redfin perch	High	13	Absent from warmer waters in Northern MDB.
brown trout	High	7	Widely distributed but restricted to cool upland streams. Stocked.

). Spangled perch, carp gudgeon complex, bony herring, and golden perch all have better estimated contemporary distributions relative to their historical records than the other native species (Table 2). Native species that we assessed as having some reduction in spatial distribution, but with relatively stable current distributions, include Murray cod, Australian smelt, and Murray–Darling rainbowfish. The remaining native species have either (a) substantially reduced distributions, (b) are not a predominantly riverine species, or (c) are poorly sampled by electrofishing, or a combination of these factors.

Of the non-native species, common carp are clearly the most successful invader and have a wider distribution than any native fish (Table 2). Eastern gambusia is also widespread, but not as well sampled by electrofishing and therefore does not demonstrate as wide a distribution as carp. Goldfish are widespread but less so than carp. The other three non-native species have restricted distributions in comparison (Table 2).

4. Discussion

4.1. Summary

The recognition of severely degraded native fish populations in Australia’s Murray–Darling Basin in the early 2000’s led to the setting up of a monitoring program to assess long-term cumulative changes in riverine fish populations at the entire MDB scale [4]. We used these large-scale data to investigate early 21st century trends in the spatial extent of 15 common native and 6 non-native riverine fish species. We found several common riverine native species displayed short-term fluctuations in extent of occurrence at the entire basin scale, but most species remained relatively stable in their contemporary distribution. All native species were collected in fewer sub-catchments compared to their pre-1980 distribution. Three non-native species—common carp, eastern gambusia, and goldfish—occurred in more sub-catchments than any single native species, but all also showed no consistent change in spatial distribution during the study period.

4.2. Species Trends

Short-lived native species generally had wider confidence intervals in their trend estimates, reflecting high year-to-year variability in their distributions, which likely resulted from their short life cycles, electrofishing sampling relative inefficiency, and a faster response to varying environmental conditions. Some smaller-bodied fish can be poorly sampled through electrofishing. Relevant to this study, carp gudgeons (CG) and flathead gudgeons (FHG) were poorly sampled in the pilot SRA (2004) where they were collected through electrofishing in 22 of 49 and 3 of 10 sites that they were known to occur in, respectively [19]. Southern pygmy perch (SPP) has an unknown susceptibility to electrofishing and is a cryptic species that favours non-riverine and/or densely vegetated habitats, and the results are treated cautiously in this riverine study. Even if there is higher sampling variability for these short-lived species, trends in estimated extent are assumed to reflect changes in actual extent, even if the extent estimates are conservative (less than actual). That is, after 10 sampling rounds, we believe that changes in estimated extent throughout time reflect true changes in extent and note that the assessments will be better understood as more years are sampled.

Carp gudgeons (CGs) are one of the most widespread and abundant taxa in the MDB and were detected in 47 of their 56 known historical sub-catchments. Only bony herring had more individual fish collected in the entire data set. Overall, we estimated that CG occurs in an average of 48% of their historic distribution in any sampling round, and because they are probably poorly sampled by electrofishing, this is quite a conservative estimate.

At the entire MDB extent, we estimated flathead gudgeon’s (FHG) current extent to be at less than 20% of its historic distribution. Although our estimate is likely to be quite conservative, it was also collected in low numbers using the multiple gear types of the NSW Rivers Survey in the 1990s [37]. FHG may be recovering in some catchments [22], but confirmation of improved population status requires more targeted or flexible sampling [38] than the current data set.

Small-bodied species that were well-detected through electrofishing as compared to other methods in riverine habitats in the pilot SRA in 2004 include eastern gambusia (collected through electrofishing in 26 of 29 sites), Australian smelt (43 of 47), mountain galaxias (12 of 15), unspecked hardyhead (16 of 21), and Murray–Darling rainbowfish (28 of 28) [19]. The pilot was conducted in dry antecedent conditions and these species may be less well-sampled under higher antecedent flow conditions. Even so, we maintain confidence in comparing relative extent between these small-bodied species.

Mountain galaxias was deemed to have a very low overall occurrence (<7% of river kilometres) relative to its known historical distribution, even though this species is known to be widespread [22]. The very low historical extent estimate may reflect lower detectability of this small species using electrofishing in lowland rivers, which have the most river kilometres in the sampling frame. In upland sites, mountain galaxias, along with other small species, is known to be particularly susceptible to predation by species such as trout and redfin [39,40]. Other factors affecting its local abundance include drought and climate change [41].

Estimates of spatial extent for spangled perch were the most variable of all common species, but the species was found in up to 80% and 75% of its contemporary distribution in the final two rounds. This increase could be associated with the species’ ability to rapidly colonise new areas following rainfall [42]. The species is common in the warmer, northern MDB and individuals have occasionally been recorded in the cooler Murray River system (Southern MDB) after extensive flooding in the northern Darling catchment [42], but they have not persisted in the Murray River.

Northern river blackfish’s contemporary extent appears stable; however, the species is extremely sparsely distributed relative to its historical distribution. There is a possibility that some of the southern historic records refer to two-spined blackfish prior to its recognition and description in 1984 [43], but the two taxa prefer significantly different habitats [22] and potential confusion is minimal. Nevertheless, anecdotally, the species has disappeared from many the larger streams that are more impaired by river regulation, barriers, and coldwater pollution. There is real concern for MDB blackfish population persistence [44,45].

Golden perch and Murray cod are subject to both hatchery stocking and recreational harvest [46]. Golden perch is a highly mobile species, operating across large riverscapes over its life cycle, showing fast responses to extensive flooding events [47], and being collected more frequently when the sampling site had above average flows in the 3 months to 3 years prior to sampling [19]. Murray cod showed a decline in the estimated spatial extent in the third sampling round, from 2010 to 2013, but was very steady from 2014 onwards. The patterns observed in these two species using our extent index are consistent with recent population abundance estimates [20] that assessed data from multiple research programs within NSW only (including sites outside the MDB). These combined results point to a recovery in distribution for Murray cod after the 1990s [20,46]. The general trends for Murray cod, golden perch, and Murray–Darling rainbowfish were also similar to those recently predicted using population models in the southern portion of the MDB [48].

Freshwater catfish’s overall spatial distribution remained fairly constant compared to its historic levels. In contrast, this species was found to be decreasing in average fish size and increasing in abundance from the 1990s through to the present by [20].

Silver perch was widespread historically but had declined over most of its range prior to the 1990s [22]. This highly mobile species has been badly affected by river regulation [49] as it relies on long stretches of river uninterrupted by weirs to maintain successful recruitment [50]. Only 9 silver perch were recorded using multiple sampling gear types in a two-year survey of 40 randomly selected sites in the NSW MDB in the mid-1990s [37]. In the SRA/MDBFS data sets, it was detected in just 21 sub-catchments across the entire 19 years, but typically only in about 20% of these in any sampling round. Compared to its historical distribution of 51 sub-catchments, it was collected in less than 10% of its historic spatial extent and in very low numbers in the northern MDB. We consider it the large-bodied native species in poorest condition relative to historical extent and it is listed as critically endangered nationally [51].

Common carp was the most frequent non-native species and estimated to occur in ~80% of MDB river kilometres. It was consistently widely distributed and was slightly more widespread in the most recent three sampling rounds compared to the first three rounds. Carp is more frequently collected when the preceding 3 months’ or 3 years’ flow levels are higher than average [19] and may be particularly more successful than the other non-native species because its populations respond to overbank flows and are favoured by some current water management practices [52,53]. The trend lines for the spatial extent indices for carp and golden perch were very similar, with both species generally increasing in extent throughout the study period but suffering slight decreases in 2014 and 2019. We suggest that this is because these two species respond in a similar fashion to flood events and flow in general, which can both operate at a basin-wide scale.

Eastern gambusia’s extent fluctuated throughout, but it is still considered a very successful invader, occurring in ~48% of river kilometres on average in our data sets. It does not migrate and is a weak swimmer, relying on flooding, drift, and rapid breeding for range expansion. This small-bodied species is more likely to be collected when the site has low flow [19], presumably as it is more concentrated within the site and the water has lower turbidity. When present, it can be locally abundant and may be reasonably well collected through electrofishing for a small-bodied species, especially in small streams. However, anecdotally, we feel that this current spatial estimate is low because it was collected in more sub-catchments than any other species. Non-detections in some large lowland streams using large boat electrofishing gear may be spurious because of its habit of occupying shallow, shoreline vegetated areas. This species is a successful invader as it is tolerant to a wide range of water temperatures, oxygen levels, salinities, and turbidities (e.g., [54,55,56]). Its local abundance declines in the winter seasons in cooler parts of the basin.

Goldfish and redfin perch were less widely distributed and their spatial extent in the MDB have remained relatively constant throughout the first 20 years of this century. Both trout species are also stable in extent and are likely limited by their low tolerance to high water temperatures [22], occurring mostly in upland and montane streams. Hence, they are unlikely to expand their distribution from their current extent naturally because of increased temperatures and reduced water availability in the MDB through climate change [10]. Nevertheless, both species are stocked and could be translocated to new catchments by angler groups.

4.3. Surveillance Monitoring Returns Coarse Assessments

Not all MDB fish species were included in the analyses because small-order streams (<5 GL average annual flow), floodplains, and wetlands were not sampled, and rare, cryptic, and/or threatened species are not well assessed by generic monitoring methods. There are 64,000 km of riverine streams in the sampling frame, and when assessing species status and trends in fish species across the whole MDB, by necessity, generalisations must be made. The restrictions on stream size and the coarse sampling method lend themselves to sampling common riverine species, which may select many species likely to be more adaptable to environmental change and habitat degradation and hence show smaller changes in trends throughout time. They are, however, also likely to be impaired by spatially extensive events, such as the drying of waterways under the expected higher frequency of extreme events predicted from climate heating, and therefore provide important data on broader extents. Furthermore, multiple common riverine species included here, i.e., Murray cod, golden perch, common carp, silver perch, and river blackfish, have been the target of multiple interventions during the past 20 years [18,28] and their spatially extensive change is of interest. Given the widespread degradation of the MDB (and many other river basins worldwide), together with investments in rehabilitation activities, data sets and analyses such as the ones used in this study provide important baselines from which to measure improvements or further declines.

On the other hand, many native fish species in Australia have evolved to respond to ‘boom and bust’ climate fluctuations (e.g., [57]) and are known responders to rapid changes in habitat. In this study, we include several species known to expand in distribution via dispersal and spawning following flood events (e.g., [3,58]), but these expansions typically occur at finer scales such as in reaches or sub-catchments. We suggest that common carp and golden perch trends in spatial extent are similar in our study because they respond in a similar manner to flow events at any spatial extent. Inevitably, when looking at spatial patterns and temporal trends both within and among species at the basin scale, there are complex and intricate influencing factors that may be considered, but most are beyond the scope of this paper. The data collected here are not aimed toward identifying causes of change but merely to identify short-term and long-term trends and overall patterns. The lack of detection of sustained long-term change in extent for common riverine species at the MDB scale only means that the cumulative effect of multiple small-scale management interventions or local impacts are not yet discernible at this extent for those species. Assessment of finer-extent objectives requires targeted follow-up surveys for specific species at finer spatial scales.

4.4. Factors That Can Influence Population Extent at a Larger Scale

Apart from the cumulative effect of local-scale interventions, factors that can influence fish populations at the entire basin scale are typically climate- and flow-related. The MDB is a highly variable, semi-arid environment that can exhibit extremes in environmental conditions, especially flow rates and water quality. For example, the Millenium Drought of 1997–2010 [59] had significant impacts on freshwater fish populations and their habitats [60], including major fish kills and deteriorated wetland extent/quality affecting many species [12]. The sampling frame for the current data sets attempts to lessen drought influences on the collection of data by restricting sampling to permanent streams. But fish assemblages in these streams remain affected by climate change, reduced connectivity, fish kills, and changes in water quality well after such events. Alternately, high-flow years provide reproductive and relocation opportunities for many flow-responsive species such as golden and spangled perches [42] and common carp. Climate change is likely to exacerbate flood and drought frequencies and magnitudes [9,41,61], and there is need for future management adaptation and ongoing monitoring [11]. This highlights the need to consider scale, environment, and sampling conditions, as well as the ecology of individual species when interpreting survey results.

Given the size of the MDB, spatially extensive scale assessments are subject to spatial–environmental variations, and in some cases further, finer-scale (e.g., river reach) or targeted, individual species analyses over smaller distributions (e.g., for trout cod Maccullochella maquariensis) [62] may be needed. This may apply, in particular, in relation to assessing rehabilitation projects that may have been conducted, or for range-limited and fragmented rare or threatened taxa. For example, such studies may be needed for assessing the distribution of migratory or highly mobile species following the installation of fishways (see [63]).

We recognise that the methods used for general, spatially extensive condition monitoring may not be equally applicable to all species or habitats [31,33,64]. The sampling methods used here were designed to be consistent over large spatial extents (in this case, 1 M km²) and to assess river kilometres, not individual fish species or fish populations. These monitoring programs are early-warning, surveillance-type programs, where there is no attempt to associate assessments to individual stressors or management actions [4]. Consequently, the methodology allows for the assessments to be representative at the MDB scale, but simultaneously, the interpretation of assessments is complex and subject to many confounding factors. Species extent estimations may vary throughout time for two reasons: (1) changes in each species true extent throughout time and (2) changes in factors that affect the calculations. Factors that could affect the true extent of each species include climatic (wet and dry years affecting flow levels), extreme event (bushfires), fishery management (stocking, harvest seasons, slot sizes, etc.), riverscape management (e.g., connectivity, habitat enhancement, or degradation) (e.g., [15,65,66,67,68]), and competition with non-native species [69]. Factors that could affect the extent estimates for each species include differences in susceptibility to electrofishing among species or different size classes (e.g., [31,64,70,71]), or antecedent conditions that influence sampling efficiency [31]. These sampling efficiency factors are of little concern with a long-term data set, however, because they do not create a bias in the trend assessments within a species or size class; (1) susceptibility to electrofishing between sizes and species remains constant—the assessments throughout time remain relative to each species, and (2) conditions also do not create a bias as they affect all species or surveys randomly across the 20 years and at the same time. In brief, a widespread, impaired sampling effect would potentially produce similar patterns in trends or changes in extent among multiple species, and this did not occur.

We acknowledge that finer-extent sampling is required to interpret finer-extent responses such as individual management actions or local climatic events. Conversely, it is clear that no common riverine species increased or decreased in its basin-wide spatial extent for a prolonged period during the study period.

5. Conclusions

We found no significant consistent trend or sustained long-term change in the spatial extent of any common riverine fish species in the entire MDB between 2004 and 2022. We believe this is a clear indication that any changes in common riverine fish species distribution are not occurring at the basin scale or for sustained periods of time.

It is important to note that these data come only from the past 20 years. The decline in many popular and well-known MDB fish species occurred well before this time period, 1950s for golden perch [72] and circa 1900 for Murray cod [73], providing an important example of the trap of ‘shifting baselines’ [74]. In that context, it should be recognised that populations and distributions of MDB native fishes were once much greater and more widespread than they are now, but we provide valuable, contemporary data over two decades for comparison with future monitoring of riverine native and non-native species via electrofishing. Across the 21 species, there were considerable differences in the variability of assessments. Species that we identified as having effect sizes > 50% of the mean should be subjected to increased sampling efforts if monitoring for changes in their extent becomes a priority. This should be expected for any monitoring program across multiple species and reinforces the need for long-term data sets to allow for a better understanding of inherent variability, in our case, in short-lived species especially.

Our results also highlight the challenges of variable population responses in highly variable environments across a large spatial scale. The detection of significant trends in highly variable ecosystems requires long-term data sets, and additional years of sampling and interpretation will add value to the existing data sets. This first 20 years of surveillance monitoring data have provided a valuable contribution to the assessment of the contemporary spatial extent of native and non-native fish populations in the MDB and have provided a baseline from which changes in the status of fishes, their protection, and population recovery or decline can occur. This is pertinent to river basins worldwide in the face of changes in hydrological regimes from climate change. The data sets use a well-designed probabilistic sample and, as such, can readily be supplemented with data collected as part of targeted monitoring programs (e.g., [75]). For example, targeted sampling (e.g., from wetland habitats, small streams, cryptic or rare species) that is informed by the ecology of each fish species may also improve the data sets. Similarly, the inclusion of other parameters that may be useful for management, such as indicators of population dynamics (reproduction, recruitment, disease prevalence, intensity, etc.) and being more predictive using population modelling (e.g., [76]), will add value.

These data and their analyses provide an important step forward to improving the management of native fishes of the MDB in Australia. The importance of long-term monitoring to guide and evaluate the benefits from the implementation of major water reforms under the Murray–Darling Basin Plan is essential. As many river basins throughout the world are under threat and have similarly reduced fish populations to the MDB, this approach is applicable to many river basins globally. Ongoing spatially extensive monitoring is also important to identify the occurrence and potential expansion of new non-native species, (e.g., oriental weatherloach Misgurnus anguillicaudatus) or to detect new incursions, especially by Tilapia species that currently occur in catchments close to the northern MDB [77].