Cold-water pollution impacts on two ‘warm-water’ riverine fish: interactions of dam size and life-history requirements

The Murray–Darling Basin

Native fish populations of the MDB are in poor condition, with overall populations now estimated to be <10% levels prior to European settlement (1850s), with the need for restoration now unquestioned (Koehn and Lintermans 2012; Koehn et al. 2020b; Murray–Darling Basin Authority 2020). There are 22 dams identified in the MDB that have the potential to cause some degree of impact on water temperatures downstream (inset, Fig. 1), with CWP reported to have major adverse effects on many river reaches (Koehn et al. 1995; Astles et al. 2003; Lugg and Copeland 2014). Most cold-water discharges occurring seasonally between September and March (austral spring to autumn) (usually involving large volume releases to supply irrigation water) and this largely overlaps with important spawning and recruitment periods for many native fish species (Koehn et al. 2020b).

Naturally, stream temperatures vary depending on meteorological conditions, location, topography, discharge, atmospheric conditions, and streambed attributes (Caissie 2006). In addition, the impact of dam releases depends on the attributes of the dam (Table 1), dam water levels, difference between surface and outlet water temperatures (greater disparity over warmer months), outlets in use, discharge volume (maximum during austral spring and summer), the downstream river channel and volume of water within the downstream reach. On occasions, dams may also spill or rapidly default to low-level outlets to prevent spilling, or to prevent the downstream dispersal of cyanobacteria (Rish et al. 2000).

Effects of CWP may be especially pronounced in the naturally warm, highly flow-variable rivers of the northern MDB. We examined the potential impacts of the following two contrasting dams in the northern MDB: Pindari Dam on the Severn River, which does have a multi-level outlet, and Copeton Dam on the Gwydir River, which does not (Fig. 1, Table 1). These two dams are considered representative for many other sites in both this region and semi-arid rivers more generally. Although both dams can affect water temperatures by ~16°C, Copeton Dam has a capacity approximately four times that of Pindari Dam, is 30 m deeper, has twice the outlet capacity and affects approximately three times the downstream river length (Table 1).

Study species

Murray cod and golden perch are widespread, important species that are highly sought after by recreational anglers (Koehn and Todd 2012). As Murray cod has a national conservation listing, it must be managed for both conservation and recreational fishery objectives (Koehn and Todd 2012), whereas golden perch has no listing. The ecology of both species is well known for the northern MDB (Koehn et al. 2020b), providing a solid knowledge base for the population models. Although both species have similar and large geographic distributions (~1 × 10⁶ km²), their life histories and reported spawning seasons and temperatures vary (Table 2, see Koehn et al. 2020b). The parameters used in this study (Table 2) were conservatively set after reviewing the existing literature and accessing recent research and the expert opinion of local researchers (Jason Thiem and Gavin Butler, NSW DPI Fisheries, unpubl. data). Proportional egg and larval survival as a function of temperature for both species were derived from Todd et al. (2005) (see Fig. S1 and S5 in the Supplementary material).

Murray cod is a demersal river species strongly associated with instream woody habitats (Koehn 2009a, 2009b; Koehn and Nicol 2014) with largely within-reach movements (Koehn et al. 2009; Koehn and Nicol 2016). Therefore, the use of a single population model for each river reach is appropriate, while inter-reach movements of all life stages in and out of each reach should balance out. Paired Murray cod adults spawn and fertilise eggs that adhere to hard substrates (e.g. hollow logs, clay banks) in slow-flowing (<0.1 m s⁻¹) areas. Eggs are guarded by the parental male until they hatch (Cadwallader 1979), with larvae surviving on large yolk-sacs for up to a week as they drift and disperse downstream (Koehn and Harrington 2006).

Golden perch is considered a flow-pulse specialist, with many components of its lifecycle driven by river flows (Sharpe 2011; Koehn et al. 2020a; Stuart and Sharpe 2020). Spawning temperature has variously been recorded up to 24°C (see Koehn et al. 2020a); however, in the model, this trigger was set conservatively at 20°C to reflect recent data (Aquatic Ecology Unit 2010; Jason Thiem, Gavin Butler, NSW DPI Fisheries). This highly mobile species has large-scale movements at all life stages, including drifting eggs and larvae (Koehn and Harrington 2005, 2006) and often considerable upstream and downstream movements of highly mobile juveniles and adults (O’Connor et al. 2005; Koehn and Nicol 2016; Zampatti et al. 2018; Koehn et al. 2020a).

Flow and water-temperature data

Daily flow and temperature data were collected from appropriate gauge locations along each river. These data were analysed within spawning seasons to estimate annual spawning rates and changes to estimates of egg and larval survival rates for each species in each population. The effects of hypoxic blackwater events (see Todd et al. 2022) were also estimated for Murray cod, and movement rates of multiple life-history stages were estimated for golden perch using flow-temperature data inputs. These estimates form the inputs for the population models. Daily water discharge and temperature data were obtained from the river gauges closest to each impoundment (upstream and immediately downstream, CWP-affected) and further downstream (not CWP-affected) of dams (Supplementary Table S1). Available flow and temperature data were downloaded from the MDBA (https://www.mdba.gov.au/) and WaterNSW (https://www.waternsw.com.au/) websites for each location. Data missing from gauges used in this study were estimated (with repetition or modelling of records from the same dataset) to ensure completeness for population modelling (see Table S1 in the Supplementary material for further details of data replication and missing value analysis).

Because of limited complete continuous-flow and temperature datasets, we used shorter but complete flow and temperature data sets to create sequences of data to be able to generate ecologically meaningful output. For the Severn River, we generated a 35-year flow and temperature data set by repeating a 7-year sequence five times (Supplementary material). For the Gwydir River, we generated a 45-year flow and temperature data set for Murray cod by repeating a 9-year sequence five times, and a 27-year flow and temperature data set for the Gwydir and Mehi rivers for golden perch by repeating a 9-year sequence three times (see the Flow and temperature data section of the Supplementary material). It is recognised that this may remove any longer-term trends (including climate change over recent decades) and may dampen natural variability over multiple decades

Population models

We reviewed previously collated ecological knowledge for each species (Koehn et al. 2020a), seeking additional opinion through expert workshops to ensure use of the most recent ecological and operational knowledge relevant to the sites. A conceptual model for the effects of CWP on fishes and their life-stages was developed for each site, assessing the likely levels of susceptibility of the different life stages to temperature impacts. The Murray cod population model was a stage–age matrix construct previously developed for the MDB (Todd and Koehn 2007; Koehn and Todd 2012) that was modified to reflect the species’ ecology in the northern MDB. Two separate cod populations were modelled (Fig. 1), including one population in the Severn River from immediately downstream of Pindari Dam to the confluence of the Macintyre River, and one population in the Gwydir River from downstream of Copeton Dam to Tareelaroi Weir.

The golden perch population models were also modified from a model previously developed for the southern MDB (Todd et al. 2022) to reflect their ecology (Table 2; Koehn et al. 2020a). The golden perch model is similar to the Murray cod model, but includes the flow-dependent movement of multiple life-stages (e.g. larvae, fingerlings, juveniles and adults) of this highly mobile species (Koehn and Nicol 2016; Zampatti et al. 2018). Golden perch populations in the Severn and Gwydir rivers were modelled with the same spatial extents as for Murray cod models above, with an additional population in the Mehi River (downstream of Tareelaroi Weir to the confluence of the Barwon River; Fig. 1). Movement (downstream larval drift, juvenile and adult movements) was estimated between the two populations on the Gwydir and Mehi rivers, making this a metapopulation model. This was necessary to model population dynamics (e.g. movement) between the two reaches split by a the instream Tareelaroi Weir that severely restricts upstream movement. Movements are also restricted by several other structures such as the Gundare Bridge Regulator and Combadello Weir. A second population was not modelled in the Severn because a natural weir at the bottom of the Severn River makes upstream movement impossible (although loss of individuals into the Macintyre River from the Gwydir population is modelled, see the Golden perch model construct section of the Supplementary material for more detail). Further details of the population-model development process, data generation, model rules and parameterisation, including specific spawning cues and flow-movement and larval drift relationships, are provided in Fig. S1–S8 of the Supplementary material.

The output from each population-model simulation was 1000 trajectories of the abundance of fish in each age-class over the period of the applied flow and temperature data for each system and scenario modelled (described in the next section). Trajectories of all adult age classes were tallied, resulting in 1000 trajectories of the total adult population size. The minimum population size from each total adult trajectory was recorded with the associated cumulative frequency distribution of minimum population sizes used to represent risk for the adult population (risk curves: Burgman et al. 1993; Todd et al. 2017). The expected value of these risk curves was also calculated to compare outputs.

CWP amelioration scenarios

To assess a range of CWP amelioration options, several temperature scenarios were explored for each species in each location. In total, six scenarios were modelled in each river, including current conditions (observed cold-water pollution as measured immediately downstream of the dams) and decreasing CWP effects by 2, 4, 6, 8 and 10°C. A counterfactual scenario was also included for the Severn River, with ‘natural’ temperature (i.e. not affected by an impoundment) taken from above Pindari Dam (data from Strathbogie gauge) combined with flow data from gauges downstream of Pindari Dam (Table S1).

Data for each cold-water amelioration scenario were generated by adding 2°C increments to daily temperature records at the site immediately downstream from dam structures in each river. As the affected river reaches are relatively short, temperature changes were not included as a gradient across the affected reaches but modelled as a uniform change. This simple CWP amelioration ‘model’ was used because examination of data on dam levels, depths of the hypolimnion and water release rules (for both irrigation water supply and to avoid releases of cyanobacteria) provided no clear pattern to predict the usage of outlets levels. This uncertainty in dam operations necessitated the use of existing datasets from gauges near the two major storages in this study.

Stocking and angler harvest

Fish stocking data for 2009–2020 were obtained from NSW DPI Fisheries for the Gwydir River reach in this study. Stocking of Murray cod in the Gwydir River was generally low (mean = 560, s.d. = 1701 across modelled years), whereas stocking numbers of golden perch were higher (mean = 7653, s.d. = 10 539 for Gwydir River; mean = 1850, s.d. = 4058 for Mehi River across modelled years). No data of stocking in the Severn River were available. To test whether stocking would assist population recovery for this golden perch population, a simulated stocking rate of 25 000 per year was added from time-step 20 (2034) onward. Stocked individuals were added into the population models as fingerlings (without penalty for being bred in a hatchery) at each relevant location. Angling harvest also occurs in all reaches; however, because data were lacking, this impact was set at an estimated conservative 5% per annum for legal-sized (adult) golden perch and 10% for Murray cod (within the legal slot size of 550–750 mm TL) (see Nicol et al. 2005; Forbes et al. 2015).

Model assumptions

For the Severn River, the average adult population size (AAPS) and initial adult population sizes (IAPS) for Murray cod were 10 000 and 10 000 respectively (assuming ~24 adults per 200 m of river length) and 50 000 and 10 000 for golden perch (assuming ~24 adults per 200 m). The models were run with a ‘burn in’ period of 50 time-steps (years) before implementing temperature effects to avoid initial population size effects.

For the Gwydir River, the density of both species was thought to be lower than in the Severn River and so the AAPS and IAPS for Murray cod were set at 10 000 and 10 000 respectively (assuming ~8 adults per 200 m of river length), with 20 000 and 2000 for golden perch in the Gwydir River (assuming ~2 adults per 200 m) and 20 000 and 5000 for golden perch in the Mehi River (assuming ~4 adults per 200 m).

Spawning opportunity analysis

Changes to temperatures are often expressed as degrees Clesius change from natural. However, there is a need to consider the effects that this may have on fish populations ecologically by affecting environmental conditions in relation to critical life stages and seasons. To provide this ecological perspective on the predicted impacts of CWP, flow and water temperature data were collated and then assessed using the spawning rules within the model for each species (i.e. days above the spawning temperature threshold). The maximum number of spawning days possible in each year under the spawning mechanism is 212 for golden perch and 92 for Murray cod (i.e. the length of the spawning season), where spawning is triggered by flow and temperature attributes for golden perch and temperature alone for Murray cod (see Fig. S1–S3 of the Supplementary material for further details). The number of days in each spawning season from all available years of data at each gauge was calculated for both Murray cod and golden perch at three locations on each river. Locations included immediately downstream of the dam, as well as an upstream and downstream location (details in Table S1). These analyses were undertaken separately to the population modelling and were designed to illustrate the likely impacts of CWP on critical spawning opportunities (and an important component within the population model), without the overlaid complexity of responses within the population models.

Spawning opportunity analysis

Examination of the daily discharge and temperature data indicated that there were considerable changes to both flows and water-temperature regimes in the rivers downstream of both dams (Fig. 2). High flows are initially reduced as water is stored, and then increase as water is released to satisfy irrigation demands. As these high-volume releases are made from low-level outlets, river water temperatures decline markedly (e.g. maximum temperatures declined from 20 to 14°C).

Fig. 2.

Plots of the daily flow (ML day⁻¹) and temperatures (°C) for (a) Severn River, and (b) Gwydir River for each study site gauge. Daily flow is shown with black lines across upper panels, and daily temperature is represented by red lines (left y-axis) in lower panels for each river. The numbers of spawning days for golden perch are included as annual horizontal black bars (right y-axis) for each site across the lower panels for each river. Note that the length of these bars spans each year and has no bearing on temporal magnitude.

Analysis of these data indicated a significant decrease in average spawning opportunities in CWP-affected reaches (Fig. 2, right-hand y-axis) compared with unaffected reaches (Table 3), up to 100% for golden perch in the reach immediately downstream from Copeton Dam where no spawning was estimated to occur across the analysed years of data. Predicted impacts of CWP on Murray cod spawning were lower than those on golden perch in the Severn River (reduced by 38% below Pindari Dam), but similar in the Gwydir River (reduced by 92% below Copeton Dam). These differences were due to the lower temperature threshold required to initiate spawning for Murray cod than for golden perch (i.e. spawning is modelled as independent of flow, see Methods). As expected, average spawning opportunities were also slightly reduced at the sites upstream of dams, owing to their higher elevation and, hence, naturally cooler water temperatures and more natural flow regimes.

Population modelling results

Severn River Murray cod

Murray cod was largely unaffected by CWP in the Severn River downstream of Pindari Dam. The population remained relatively stable over time, with little distinction among the different scenarios tested (Fig. 3, Table 4). Remediating CWP had little effect on Murray cod, with only the counter-factual scenario completely removing any cold-water effects (Fig. 3a, b), with negligible percentage differences from the baseline CWP scenario (Table 4). Spawning occurred every year as indicated by spawning analysis (Fig. 2, Table 3). Every 7th year, because of the constructed nature of the flow-temperature timeseries, Murray cod larvae were strongly affected by CWP, although there was recruitment to 1-year olds every year. Including 10% angler harvest further decreased population outcomes with no additional CWP effects (Fig. 3c, d).

Fig. 3.

Murray cod average adult population size for the Severn River with (a) cold-water pollution (CWP) effects and remediation scenarios; (b) associated risk curves; (c) CSP remediation scenarios with fish harvest; and (d) associated risk curves. CWP-affected temperature scenario increases (a, b) are indicated by coloured lines in the key (+°C). Fishing (c, d) includes angler harvest.

Table 4.Expected minimum population sizes for Murray cod and golden perch for the Severn and Gwydir–Mehi rivers, with water temperatures changed by +2 to +10°C from the cold-waterpollution (CWP) scenarios modelled.

Item	CWP	CWP + 2°C	CWP + 4°C	CWP + 6°C	CWP + 8°C	CWP + 10°C	CF
Severn River
Murray cod	7175	7127	7150	7192	7168	71 668	7413
Murray codA	5757		5808
Golden perch	0.58	122	300	256	271	267	3.08
Golden perchB	22		569
Gwydir River
Murray cod	826	1717	3781	5785	6484	6490	n.a.
Golden perchC	0.8	0.8	1.5	8.2	129.1	130.0	n.a.
Golden perchD	12.1	12.2	20.5	71.3	333.2	336.7	n.a.
Golden perchE	150	147	160	229	1078	1089	n.a.
Golden perchF	4020	3965	3978	4117	4770	4804	n.a.

The expected minimum population size is the expected value of the distribution of the minimum population size collected from each scenario trajectory, i.e. the expected value of the risk curve.

CF, counter-factual scenario, with ‘natural’ temperature taken from above the impoundment.

^AWith fishing.

^BWith stocking.

^CGwydir with barriers.

^DGwydir without barriers.

^EMehi with barriers.

^FMehi without barriers.

Severn River golden perch

Golden perch was significantly affected by CWP in the Severn River downstream of Pindari Dam. The modelled population declined to near-extinction for the CWP scenario, but was predicted to improve when CWP was remediated (Fig. 4, Table 4). Remediating CWP improves the outcomes for golden perch, with a 4°C increase producing the best results. However, similar improvements were predicted for 6, 8 and 10°C increases, with the percentage differences from the baseline CWP scenario being substantially increased for a 4°C increase and higher (Table 4). Recruitment to 1-year-olds largely occurs every 5 years (Fig. 4a, b), again because of constructed nature of the flow-temperature timeseries, with the greatest recruitment events occurring with 4, 6, 8 and 10°C increases. Even with the improved outcomes from at least a 4°C increase, there is still a high likelihood that the population will be small (Fig. 4). However, stocking may mask the effects of CWP and, in conjunction with remediation, may stabilise a golden perch population (Fig. 4c, d).

Fig. 4.

Golden perch average adult population size for the Severn River with (a) cold-water pollution (CWP) effects and remediation scenarios; (b) associated risk curves; (c) CWP effects and remediation scenarios with annual stocking of 25 000 fingerlings beginning in 2034; and (d) associated risk curves. CWP-affected temperature scenario increases (a, b) are indicated by coloured lines in the key (+°C). Fishing (c, d) includes angler harvest.

Gwydir River Murray cod

Murray cod was significantly affected by CWP in the Gwydir River downstream of Copeton Dam, with severe population declines and generally low population sizes (e.g. CWP scenario had ~50% smaller population size than did CWP + 2°C) (Fig. 5a, b, Table 4) without any CWP remediation. Remediating CWP in the Gwydir River improved the modelled outcomes for Murray cod, with 8 and 10°C increases producing the best results (i.e. ~680% increase in population size above CWP scenario), although gains are also made with a 6°C increase (Fig. 5a, b, Table 4).

Fig. 5.

Murray cod average adult population size for the Gwydir River with (a) cold-water pollution (CWP) effects and remediation scenarios; (b) associated risk curves; (c) CWP effects and remediation scenarios on 1-year-olds; and (d) associated risk curves. CWP-affected temperature scenario increases (a, b) are indicated by coloured lines in the key (+°C). Fishing (c, d) includes angler harvest.

Gwydir River Golden Perch

When modelling golden perch as a metapopulation with no barriers to movement, there were two levels of responses to CWP, based on whether there were barriers to movement or not and which population was being assessed (Fig. 6, 7). There were significant effects of CWP with and without barriers in the Gwydir River (Fig. 6), whereas the effects of CWP were marginally different when barriers were removed in the Mehi River (Fig. 7, Table 4). Remediating CWP in the Gwydir River improves the outcomes for golden perch, with 8 and 10°C increases producing the best results. There is an important change in population response with the removal of barriers; however, CWP remains a significant threat for smaller temperature improvements (2–6°C) and it is only the higher levels (8°C+) of CWP remediation (8°C+) that further population increases and reduced risks occur (Fig. 6, Table 4).

Fig. 6.

Golden perch average adult population size for the Gwydir River with (a) cold-water pollution (CWP) effects and remediation scenarios; (b) associated risk curves; (c) CWP effects and remediation scenarios with barriers removed; and (d) associated risk curves. CWP-affected temperature scenario increases (a, b) are indicated by coloured lines in the key (+°C). Fishing (c, d) includes angler harvest.

Fig. 7.

Golden perch average adult population size for the Mehi River with (a) cold-water pollution (CWP) effects and remediation scenarios; (b) associated risk curves; (c) CWP effects and remediation scenarios with barriers removed; and (d) associated risk curves. CWP-affected temperature scenario increases (a, b) are indicated by coloured lines in the key (+°C). Fishing (c, d) includes angler harvest.

Golden perch in the Mehi River is not directly affected by CWP. However, because of larval drift and other movement rules, the Mehi golden perch population is dependent on the Gwydir population. This dependency is captured in the modelling results where, when modelled with barriers, there is a significant effect on golden perch because of CWP, even though the Mehi River is not affected by CWP (Fig. 7a, b, Table 4). Remediation of CWP, with existing barriers in place occurs only with an increase of 8°C or more. However, with barriers removed, the CWP effect on golden perch populations in the Mehi River is substantially reduced, with much smaller temperature improvement CWP scenarios (Fig. 7c, d, Table 4), with much greater predicted population improvements and significantly reduced risks to the population.

Interactions of species ecological requirements, dam size and CWP impacts

Golden perch differs from Murray cod in having a higher temperature requirement for spawning and greater mobility throughout its life stages (and, hence, a need for extended riverine connectivity). Analysis and interpretation of the likely ecological impacts of the changes to water temperatures proved useful to highlight the effects on fish population processes. There were significant reductions of the days available with appropriate conditions for spawning within the defined seasons for each species. While spawning is not the only important factor included in the population modelling, it is an important factor, and reductions in spawning opportunities do negatively affect populations. Although the effects of CWP had not previously been explored for golden perch, it was anticipated that the effects would be greater for golden perch than for Murray cod. Indeed, temperature requirements for spawning and recruitment of golden perch (Koehn et al. 2020a) are higher than for many other, but not all, species present at these sites and hence CWP remediation would also meet the needs of other species with similar or lower temperature needs, such as silver perch, Bidyanus bidyanus (Tonkin et al. 2019), or Murray cod. The 20°C threshold used for golden perch spawning was considered somewhat conservative and the effects would be expected to be greater if this key temperature were higher, as has been previously suggested (20–24°C, see Koehn et al. 2020a). Higher temperatures may also be needed for other species such as freshwater catfish, Tandanus tandanus (>20 or 24°C, see Koehn et al. 2020a), and it is best that ecologically significant temperatures for such species be used as targets for remediation.

The CWP-modelled impacts from Pindari Dam on golden perch in the Severn River showed that the population will decline, with a very high likelihood of the population being small if no remediation is undertaken. Our modelled results indicated that population improvements are somewhat attenuated after an increase of ~4°C. The attenuation possibly occurs because the temperature increases above 4°C may modify the start of the spawning season, from which no additional benefits occur because of being confounded with the constructed nature of the flow-temperature timeseries. It is unlikely that this would occur under ‘real world’ circumstances, and although our modelling did not indicate it, it is likely that there would be further improvements beyond a 4°C increase in water temperature. Although it is possible that this temperature sequence may have underestimated the CWP effects on Murray cod, the overall outcome is clear in that CWP poses a major threat to golden perch populations, and a lesser one to Murray cod in the Severn River. CWP also has a major impact on the population of Murray cod in the Gwydir River downstream of Copeton Dam. Reinstatement of a natural temperature regime would dramatically increase the adult Murray cod population size (84% over 45 years) and decrease risks of population reductions. This modelled response to CWP remediation effects is much greater than those for the Severn River.

The difference in modelled CWP impacts between the two dams in this study is at least in part due to their size, with greater effects from the larger Copeton Dam than from the smaller Pindari Dam. A priority may be to remediate the larger dams to give the greatest environmental benefits. The impact of CWP is not only a result of the dam size and location, but also of factors such as dam fullness, the difference between surface and bottom water temperatures (typically greatest in warmer seasons), the low-level outlets that are in use, volume of discharge and other operational rules such as not releasing cyanobacteria or supplying drinking water (Rish et al. 2000; Weber et al. 2017). The impacts form Pindari Dam may also have been ameliorated by its multi-level outlet, which would provide warmer water when operated correctly. However, full use of this outlet is often limited owing to cyanobacterial blooms making its efficacy hard to determine. Such complexities result in site-specific effects and indicate the necessity to assess CWP impacts on a case-by-case and species-specific basis.

Fish passage, angler harvest and stocking

There are many other impacts on native fishes that may either compound the effects of CWP or prevent population recovery through remediation measures that need to be considered (Koehn and Lintermans 2012). In particular, these include the need for fish passage (Baumgartner et al. 2014), adequate river flows and stocking and harvesting for recreational angling species (such as Murray cod and golden perch). Our modelling in the Gwydir–Mehi River system highlighted the importance of fish passage for a highly mobile species such as golden perch, indicating the need for considering CWP in conjunction with barriers to movements. Our exploration of angler harvest within the slot legal limit at the modest rate of 10% also illustrated the potential additional impact that additional fish removal may have on population recovery. Here, simulated harvest almost completely negated the positive effects of 4°C of CWP amelioration, likely as a function of fewer adults, and hence reduced reproductive potential of the population. Overall, this suggests that angling is likely to dampen the positive influence of CWP amelioration efforts. For example, when the angler harvest is included, there is a further reduction in populations. CWP has a much greater impact on golden perch where recruitment to 1-year olds shows episodic recruitment, with only two recruitment events in the 7-year repeated cycle modelled. Conversely, stocking may mask the effects of CWP through population stabilisation, but will not necessarily restore population function. However, stocking did improve ‘recruitment’ for some populations, which would be unlikely to remain extant without continued stocking.

Remediation options

Remediating CWP is possible through a range of engineering and operational options (Sherman 2000; Michie et al. 2020c). For example, multi-level outlets have been successful in mitigating thermal pollution (Du Xiaohu 2009; Wu et al. 2011). Whereas retrofitting older dams may be costly (Sherman 2000), there are other infrastructure options such as trunnions, destratification mixers, bubble mixers and thermal curtains that divert warmer surface waters to low-level outlets (He et al. 2017; Gray et al. 2019; Michie et al. 2020b). The utility and effectiveness of these options will be site-specific, and improvements to water temperatures and their potential benefits require quantification.

Although our results indicated that remediation of CWP could help restore populations of Murray cod and golden perch, even at dams with selective withdrawal infrastructure, full remediation of CWP may still not occur because of structural or operational limitations (e.g. Acaba et al. 2000; Rish et al. 2000). The delivery of irrigation flow water is largely dependent on the seasonal demands of farmers, which unfortunately overlaps considerably with the spawning and critical early life-stage seasons of many MDB warm-water fishes. Avoiding the delivery of cold water in these key seasons would be helpful as an interim measure. In the absence of this option, remediating CWP largely relies on investment in engineering and operational options (Sherman 2000; Boys et al. 2009; Michie et al. 2020c). It is especially important that CWP mitigation options are included in the design and construction of new dams or the expansion of existing dams. There may also be some operational alterations, such as water delivery times, that could assist in reducing the downstream impacts, particularly in years when delivery of irrigation water is limited (Tonkin et al. 2023). Additional to any planned remediation or construction of new dams, the design must predict daily temperatures and these temperatures need to be modelled to ascertain their appropriateness for the species affected.

The existing low abundances of MDB fish populations estimated to be <10% of pre-European levels (Murray–Darling Basin Ministerial Council 2004; Koehn and Lintermans 2012; Murray–Darling Basin Authority 2020) may also be an impediment to any rapid improvement in fish populations. Data from recent fish surveys in the northern MDB (from the Sustainable Rivers Audit and Murray–Darling Fish Surveys; Greg Ringwood, MDBA, unpubl. data) indicated low existing population abundances for both species, which would mean long recovery times, even with threat remediation. Additional measures such as stocking or additional harvest restrictions may be needed to facilitate recovery. Hence, population restoration is more complex than just remediating any one threat, and requires an holistic approach.

Implications for water management

Given the widespread nature and serious impacts of CWP, it also potentially compromises the recovery of fish populations through other rehabilitation programs such as the use of environmental water (Arthington 2012), which includes the Basin Plan, costing ~A$13 billion (Murray–Darling Basin Authority 2011). Most attention in such water delivery is placed on the quantity of water available, rather than its quality. Without providing the temperature thresholds needed for the adequate completion of fish life-cycles and recruitment into adult populations, the objective of this and other environmental programs (e.g. habitat re-instatement) may not be achieved and expenditure may be wasted. This places the remediation of CWP as a priority action in many rivers. The requirements of golden perch have been used as an example for species in this study, but by using ecologically significant threshold-temperature requirements relating to the needs of other aquatic species will provide the best outcomes from remediation.

Benefits, constraints and future advances of population modelling

Population modelling provides an important method by which to evaluate the likely impacts of threats such as CWP. It also provides a mechanism to assist managers to prioritise rehabilitation options to improve native fish populations and to ensure the best value for expenditure (Koehn and Todd 2012). Modelling of CWP has been piecemeal, and a strategic and consolidated approach across other sites and species would help further quantify this very widespread threat. Ecological knowledge forms a strong basis for such modelling, and this knowledge is now generally very good for Murray cod and golden perch, the subject species for this study (Koehn et al. 2019, 2020a). However, it is recognised that this may not be so for other locations and species.

Modelling in this project was hampered by the availability of complete long-term flow and water temperature datasets, and this highlights the need for more comprehensive, consistent and accurate water-temperature monitoring with the ready availability of such data. Temperature underpins many physiological and ecological functions of fish (Johnston and Dunn 1987; Whiterod et al. 2018) that are not considered within this study. For example, CWP affects swimming capabilities (Lyon et al. 2008; Whiterod 2013) and may pose a barrier to fish movement, which has wider implications for predator avoidance and the movements, recolonisation and population dynamics of mobile species such as golden perch (Koehn et al. 2020a). Temperature is also an important regulator of riverine processes such as productivity and the life history of invertebrates, an important food source for fish.

Population modelling is needed for other affected river reaches and a range of other species and life-stages, especially threatened species such as silver perch and the many small-bodied fishes that are often forgotten (Todd et al. 2017). Early life-stages are often the most vulnerable (Raymond et al. 2022), but often receive the least research attention (Koehn et al. 2019). There is also a need to identify biological, ecological and limnological knowledge gaps and undertake relevant research (e.g. minimum site-specific ecological temperature tolerances) to inform operational targets and protocols, and the development of monitoring and evaluation plans (Mejia et al. 2020).