Risk v. reward for responsive movements by a highly mobile fish species in a flow-constrained and barrier-laden river

Study area

The Gwydir catchment located in north-western New South Wales (NSW), Australia, is one of nine major catchments within the NSW portion of the MDB (Fig. 1). The study area was across the lowland sections of the Gwydir catchment and encompassed the upper Mehi and lower Gwydir rivers (Fig. 1). In the Mehi River, the study reach extended from Tareelaroi Weir, where the Mehi diverges from the Gwydir, downstream to the township of Moree (Fig. 1). In the Gwydir River, the study reach extended from 6 km upstream of Tareelaroi Weir, downstream into the Gwydir and Gingham watercourses below their confluence (Fig. 1). Typically, the river channel within both rivers is less than 25 m wide and 3 m deep. The Mehi is generally narrower (<15 m) and has the steeper gradient of the two systems. The Gwydir catchment is considered highly regulated, and the natural variability and the volume of water that now passes through its various rivers and streams is much reduced and highly controlled (Copeland et al. 2003). Through its two main tributaries, namely, the Gingham Watercourse and Gwydir River lower arm (Fig. 1), the Gwydir River terminates in a series of ephemeral wetlands that only connect to the mainstem of the much larger Barwon River during extreme flood events. Conversely, the Mehi River, although largely reliant on the diversion of water from the Gwydir River to flow, connects more often to the Barwon River, but fish passage is made difficult by the changed hydraulics the system now experiences and the presence of several large man-made instream barriers present along its length. Land use across the Gwydir catchment is primarily intensive agriculture, with the land adjacent to both rivers used mainly for irrigated crops. The system receives irrigation, stock and domestic, and environmental water releases from Copeton Dam (29.900742°S, 150.927682°E; capacity, 1346 GL), ~200 km upstream of the study area.

Fig. 1.

Acoustic telemetry array layout in the Gwydir (upper and lower), Mehi and Gingham rivers used to monitor the movement of golden perch (Macquaria ambigua). The locations of the larger weirs and regulators are also illustrated.

The instream habitat of both systems comprises a diversity of meso-habitats, including woody debris, undercut banks, overhanging riparian vegetation and aquatic macrophytes. The lower Gwydir traditionally supported populations of a range of native fish species, including notable large-bodied species (growing to >100 mm) such as freshwater catfish (Tandanus tandanus), Murray cod (Maccullochella peelii) and golden perch. Populations of all three species in the lower Gwydir are now in a poor state (Crook et al. 2023) and golden perch populations specifically are thought to be largely dependent on the stocking of hatchery-reared fingerlings to maintain local populations (Commonwealth Environmental Water Office 2019).

Fish collection and tagging

The intention at the outset of the project was to collect and tag a total of 25 ‘resident’ golden perch from within the confines of the study area, ideally with numbers similar from each of the Gwydir and Mehi rivers. However, after extensive sampling on 9–11 February 2021 using boat-mounted electrofishing (2.5-kW Smith-Root Model GPP 2.5 H/L), only 11 resident golden perch of suitable size were captured and tagged in the two rivers combined (Mehi, n = 2; Gwydir, n = 9). To meet quota, on 19–20 April 2021, gill nets (mesh size 75–100 mm) were used to source an additional 14 golden perch from upstream in Copeton Dam for translocation and release into the study area. Previous studies involving the translocation of golden perch in the Broken River (Crook 2004) and freshwater catfish from Copeton Dam to the lower Gwydir (Carpenter-Bundhoo et al. 2020) reported little or no difference between the behaviour of resident and translocated fish after an initial settling period. Copeton Dam fish were held overnight in a 4-m³ floating cage, before being transported in aerated drums (220 L) to riverine sites the following day.

All riverine fish were tagged and released on the day of capture, whereas those from Copeton Dam were tagged and released on the day of transportation. Prior to tagging, fish were anaesthetised in ambient water containing 50 mg L⁻¹ benzocaine (ethyl-p-aminobenzoate) (Sigma–Aldrich, Shanghai) and weighed (g) and measured (mm). Fish were then transferred to an operating cradle, with water containing an equivalent level of anaesthetic (50 mg L⁻¹) continually pumped over the gills to maintain anaesthesia. To access the peritoneal cavity, an incision was made through the body wall of the fish, adjacent to the linea alba and anterior of the anal vent. The sex of the fish was determined by attempting catheterisation of the fish’s vents as described by Ogburn et al. (1994) for silver perch (Bidyanus bidyanus), where the catheter can be easily inserted into the oviduct of the female with minimal pressure but is not able to be inserted into the male. Acoustic tags used were Innovsea V9 69 KHz acoustic transmitter (delay 80–160 s, approximate battery life of 859 days), which conformed to the recommended maximum tag weight of 2.25% of bodyweight for all fish (Jepsen et al. 2002; Butler et al. 2009; Wagner et al. 2011). Incisions were closed with two or three sutures using 0.3-mm pseudo-monofilament, absorbable thread (Vetafil Bengen; WdT, Garbsen, Germany). After suturing, the fish were returned to an aerated tub to recover before being released within 10–20 min.

This project was undertaken following the guidelines of and in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes (NSW Department of Primary Industries permit number 14/10).

Acoustic array

The movement of tagged fish was monitored by a linear array comprising 40 Innovasea VR2W 69-kHz receivers that covered ~180 km of the main river channels across the lower Gwydir catchment (Fig. 1). For visualisation and analysis purposes, the array was broken into four subarray sections: the Mehi River array, the Gingham Watercourse downstream of the Gingham Weir array, the Gwydir River lower arm downstream of the Tyreel Weir array, and the Gwydir River array, which extended from above Tareelaroi Weir to the junction of the lower arm and Gingham Watercourse (Fig. 1). The Mehi array spanned 31 km using 13 receivers, the Gingham array spanned 72 km using 8 receivers, the lower Gwydir array spanned 43 km using 4 receivers, and the upper Gwydir array spanned 37 km using 15 receivers. The receivers recorded binary presence or absence data, with the presence of a tagged fish logged when it entered the reception range of a given receiver (>50 m).

Statistical analysis

River path spatial objects were created using current satellite imagery of the study reaches in QGIS (ver. 3.20, see www.qgis.org). A linear distance matrix for acoustic receiver locations was then generated using Euclidean distances along the river paths in the V-Track package (ver. 2.2, see https://github.com/RossDwyer/VTrack; Campbell et al. 2012) in R (ver. 4.2.2, R Foundation for Statistical Computing, Vienna, Austria, see www.r-project.org) and the distance between individual fish detections was calculated. A daily sum of total distance moved, including both upstream and downstream movements, was then calculated for each fish. The first 24 h of detection for each fish was removed from analysis to discount any abnormal behaviour resulting from the translocation and tagging procedures (Carpenter-Bundhoo et al. 2020). A binary movement response was calculated as a detection of an individual on a minimum of two receivers within a given 24-h period for movement (1) or, otherwise, no-movement (0). In this daily movement database, individual fish were also assigned a no-movement (0) on days where they were not detected if the day was not after the final detection for that given fish, and the last and next detections were not from terminal receivers for that given fish. Otherwise, non-detected days were omitted from the database. Total linear range was calculated as the river distance between the most upstream and downstream detections for each fish over the study period, and cumulative distance moved was calculated as the sum of all upstream and downstream movements for each fish for the study period.

Generalised additive mixed models (GAMMs) fitted with binomial distributions and a logit link were used to analyse probability of movement. All models were built using the mgcv package (ver. 1.9-1, S. Wood, see https://cran.r-project.org/package=mgcv/; Wood 2003). Total daily river discharge, antecedent flow, river (Gwydir or Mehi), day of year (DOY), mean daily air temperature, lunar illumination, and fish length, weight and sex were all included as candidate independent variables. Source (river or impoundment) was also considered as a candidate variable but was not found to be a significant predictor of either likelihood of movement or distance moved and was removed during model selection. Antecedent flow was calculated as a rolling sum of the previous 30-day river discharge, generated from flow gauges relevant to an individual’s current position in the river system. Lunar illumination (%) was generated using the R package lunar (ver. 0.2-1, E. Lazaridis, see http://statistics.lazaridis.eu). Discharge and antecedent flow were transformed with a natural logarithm because they were highly skewed. A random intercept was included for each transmitter (ID) in the models because the data comprised repeated observations of each tagged fish. Year of study (Year) was included as another potential random intercept to account for yearly variation in environmental parameters not accounted for by the fixed effects. Continuous predictors were fitted with cubic splines, DOY was fitted with cyclic cubic splines, and river was included as a parametric term. An autocorrelation structure (AR1) was incorporated in each GAMM to account for temporal autocorrelation. All possible models and interactions were examined, and the best model was selected using Chi-Square tests of restricted maximum likelihood (fREML) scores. Following the protocol of Zuur et al. (2010), data were checked for outliers and collinearity among predictor variables was assessed using variance inflation factors (VIF) in R.

Within catchment movement

Region- or subcatchment-specific population processes (e.g. nurseries, source–sink dynamics) are important in understanding golden perch ecology (Zampatti and Leigh 2013). In the southerly extremities of its distribution, golden perch movement has been reported to be constrained by low-head weirs used to control irrigation deliveries (e.g. O’Connor et al. 2005; Koster et al. 2014, 2017), and weir pools have also been shown to lessen the effects of river-level rises, resulting in the reduced frequency of large-scale movements (Koster et al. 2020). In addition, not unlike the movement predicament the species experiences in the Gwydir system, golden perch in the Moonie River in the far northern MDB were confined to waterholes for extended periods because of a lack of connectivity along the main channel (Marshall et al. 2016). These studies have parallels with the behaviour of golden perch in the current study and reiterate the effect river regulation is having on the species across the entire MDB.

Although most river systems across the MDB share similarities in how they are regulated, the surface-water connectivity of the Gwydir system is also somewhat unique from other locations where golden perch movement has been studied (e.g. Llewellyn 2014; Koehn and Nicol 2016; Koster et al. 2017; Zampatti et al. 2018; Thiem et al. 2020). The substantial ‘raft’ of large wood debris that spans over 12 river kilometres in the lower Gwydir River acts as a substantial in-channel barrier for water and fish (Copeland et al. 2003; Erskine et al. 2012) and has no equivalence elsewhere in the MDB. The Gwydir Raft is an artefact of upstream land-clearing activities that commenced in the 1800s (Copeland et al. 2003), and has resulted in considerable stream braiding and shallow-bypass channelling in its vicinity both upstream and downstream (Erskine et al. 2012). Only three tagged fish traversed the raft in our study, and this occurred only on the highest discharges recorded during the study (Fish 51060, Fish 51063, Fish 51076; Fig. S1). The combination of water management and the raft is preventing migratory fish such as golden perch from transiting the catchment and the flow-cued response of adult fish is also leading them to ephemeral waterways, which are both clear illustrations of the impacts of HIREC and highlight how what was once a low risk now has no reward (Sih 2013).

Although our model did not predict a difference in golden perch movement between the Mehi and Gwydir rivers individually, there was a strong difference in movement responses in relation to discharge between the two rivers. Although being physically linked, the Gwydir and Mehi rivers have distinct flow regimes, with the Gwydir receiving far higher flow magnitudes because of differences in the amounts of irrigation water it receives and due to the different flood-mitigation strategies employed for the two systems. A similar disparity in movement responses between the two rivers was observed in Murray cod (Maccullochella peelii) and freshwater catfish (Tandanus tandanus) (Carpenter-Bundhoo et al. 2023), with both species being more likely to move with increased discharge in both systems but there was a higher probability of movement on smaller discharges in the Mehi River. The difference in golden perch responses at the river scale most likely reflects alternative movement strategies for differing hydrologic habitats. In the Gwydir River, where higher-magnitude and more frequent flows occur, golden perch exhibited an almost linear movement response to increased discharge. However, in the Mehi River, golden perch displayed a more bimodal response in movement strategies, opportunistically moving as connectivity increased on smaller discharges (as per Carpenter-Bundhoo et al. 2023), and then moving again only on high-magnitude discharges. This disparity in behaviour highlights the need to identify and consider the specific hydro-ecological responses of fishes and the characteristics of the rivers they inhabit when developing water-management strategies for a region.

Despite the numerous high-discharge events and flooding experienced during the current study, there was no evidence of simultaneous large-scale movement by golden perch indicative of a synchronised spawning migration. This contrasts the telemetry study of O’Connor et al. (2005) in the mid-Murray River system, where golden perch moved unidirectionally in spring, before aggregating below the confluence of rivers. Golden perch in the Goulburn River acted similarly, moving downstream in spring on elevated environmental-flow releases, with spawning being confirmed by larval drift net sampling (Koster et al. 2017). The increase in the activity by some golden perch individuals from September through to December in the current study does fit the perceived window of spawning for the species (e.g. O’Connor et al. 2005), but was largely individualistic and coincided with an increased flow, and is likely to have represented opportunistic relocations rather than spawning-related movements (i.e. no synchronised movements or aggregations of tagged fish). Unlike their southern conspecifics, golden perch in the northern MDB have been reported as spawning across an expansive window within an annual cycle (Ebner et al. 2009; Thiem et al. 2023). However, the Darling River and its tributaries now support suitable conditions for system-wide spawning and recruitment only during extended and consistent periods of high flows (e.g. La Niña conditions), which occur infrequently across the broader northern MDB and even more rarely in the Gwydir (Stuart and Sharpe 2020; Thiem et al. 2023).

The perceived poor state of the golden perch population in the Gwydir system is supported by the findings of several surveys undertaken in recent decades, with only low numbers captured in all years sampled and negligible evidence of natural recruitment (e.g. Commonwealth Environmental Water Office 2019; Crook et al. 2023). Collectively, this suggests that the Gwydir now functions as neither a source nor a sink for golden perch within the broader context of the MDB (Fig. 4). Effectively, it now appears to be more akin to terminal wetland system like the Lachlan River located in the southern MDB, which, like the Gwydir, has little or no migration of golden perch into the system and lacks any substantial natural recruitment (e.g. Shams et al. 2020; Schilling and Crook 2023).

Fig. 4.

A conceptualisation of managing a false-sink population of golden perch (Macquaria ambigua) in the Gwydir River system. Shaded grey boxes represent the findings of the current study.

Migration and immigration

The Gwydir system is a highly modified riverscape that is effectively unrecognisable in the context of what could be considered a naturally flowing and unimpeded system (Yarnell et al. 2015). The combination of the ‘barrier maze’ created by myriad instream anthropogenic infrastructure and the Gwydir raft, and the resulting modified flow regime, has resulted in the Gwydir system transitioning to become a largely isolated compartment of the MDB. In addition, Copeton Dam is one of the largest instream dams in the MDB and, like large headwater impoundments globally (Barbarossa et al. 2020; Belletti et al. 2020), has also profoundly affected the flow regime across the entire system below. Water quality is also affected for vast distances below the dam, including cold-water plumes that extend 100 km into the systems lower reaches (Copeland et al. 2003). The combined impact of these environmental changes on the broader riverscape means that the fish community in the Gwydir no longer functions as it once did.

The results of the current study suggest that movement of adult golden perch between the Gwydir system and the broader MDB population is limited at best. This was not the case historically, with a healthy population of non-supplemented golden perch present throughout the system prior to river regulation (Copeland et al. 2003). The construction of Copeton Dam has seen the virtual loss of natural winter flow peaks, an increase in smaller flow events in summer because of irrigation releases, and the almost total absence of large floods (the latter originally the main means by which the Gwydir connected to the Barwon River) (Thoms et al. 2001). In this study, several golden perch exited at the downstream extent of the array in the Gwydir and Gingham channels in a possible attempt to move to the Barwon River, but the majority returned to the upstream reaches sometime after or were presumably trapped in the wetland below, due to there being insufficient flow to fully connect the Gwydir and Barwon rivers.

The degree of riverscape modification that has occurred over the past century in the Gwydir catchment appears to be too great to allow a self-sustaining population of golden perch to perpetuate. River regulation and barriers have led to a loss of pelagophils in arid and semi-arid rivers globally, mainly due to the loss of flow-related life-history cues, the lack of stream connectivity constraining spawning aggregations, and that there are now insufficent stretches of uninterrupted river channel in many systems to support drifting eggs and larvae (e.g. Dudley and Platania 2007; Hoagstrom and Turner 2015; Lechner et al. 2016). Furthermore, a decline in stream invertebrate productivity, as a further consequence of regulation, has rendered many rivers incapable of sustaining drifting pelagic larvae and juvenile pelagophiles (Balcombe et al. 2005; Humphries et al. 2020). Effectively, HIREC means that the life-history strategies of pelagophils have become largely ineffective in highly regulated rivers like the Gwydir. Juvenile and subadult golden perch immigration is one possible means by which the Gwydir population could be bolstered, but requires intimate knowledge of golden perch dynamics and the reinstatement of local connectivity (e.g. Stuart and Sharpe 2020; Stuart et al. 2024).

Empirical information is critical to understanding source–sink dynamics (Diffendorfer 1998). Whether or not the Gwydir catchment historically contained a source or sink population of golden perch is unknown. Results from the current study suggest that adult golden perch are now unlikely to move between the Gwydir and Barwon river systems even in wet years. Under the current scenario, a better management strategy to promote golden perch movement (including adult and subadult movement) may be to work toward reconnecting the Mehi River through targeted flow delivery, rather than the assumed traditional pathway via the Gwydir wetland complex (Fig. 4).

Flow indicators

Functional groups of fishes have been identified and used in monitoring, prescribing, and managing environmental flows across the MDB, and golden perch has been a prominent species used as a key indicator species in this regard (King et al. 2009; Zampatti and Leigh 2013; Baumgartner et al. 2014; Stuart and Sharpe 2020; Stuart et al. 2024). However, our study indicated that golden perch is not currently a useful flow indicator species within the Gwydir system. Species that respond to localised changes in flow delivery in-channel or on floodplains are likely to be better indicators and potentially should be given precedence when designing managed flow releases.

Baumgartner et al. (2014) recognised flow-dependent specialists and floodplain specialists as two important functional groups in a review of the southern MDB. Of the two flow-dependent specialists recognised in that study (golden perch, silver perch), neither exists as self-sustaining populations in the Gwydir system. Small-body in-channel species such as the Darling River hardyhead (Craterocephalus amniculus) (Moy et al. 2018), or floodplain specialists such as olive perchlet (Ambassis agassizii), both of which can complete their lifecycle at a smaller extent (e.g. reach scale), may be better environmental indicators in the Gwydir than are larger-bodied ranging species such as golden perch. Although the concept of using small-bodied species as flow indicators is sound, society rarely rallies behind small non-iconic fishes (Ebner et al. 2016). As such, at least one large-bodied, non-rheophilic species, with some level of flow dependency (e.g. freshwater catfish), is likely to be a better focal species for targeted environmental flow releases in the Gwydir system (Carpenter-Bundhoo et al. 2020, 2023). Alternatively, the fall-back position is to monitor species that have shown some level of response to environmental flows in recent decades in the Gwydir and elsewhere in the MDB (e.g. spangled perch (Leiopotherapon unicolor) and bony herring (Nematalosa erebi) Wilson et al. 2010). For the foreseeable future, the ongoing put-and-take fishery for golden perch (Fig. 4) will remain important to local anglers and, by extension, for maintaining relevance for fish and river management for local people. To rehabilitate the fish community in the Gwydir system, and especially to recover flow-dependent specialists, major institutional and societal change at all levels of government and the wider community alike is required (Fig. 4) (Johnson et al. 2017; Claussen and Philipp 2023; Terui et al. 2023).