Seabirds are important bio-indicators within the marine environment (e.g., Croxall et al., 2002; Lascelles et al., 2012) and are currently one of the most threatened group of birds, with 110 of 359 seabird species threatened globally (reviewed in Dias et al., 2019). Seabirds are highly sensitive to changes in their food resources (Reid and Croxall, 2001; Crawford, 2007; Frederiksen et al., 2007; Chiaradia et al., 2010), and disappearance of their primary prey often result in decreased breeding success and population size (Croxall et al., 1999; Dann et al., 2000; Lynnes et al., 2004; Österblom et al., 2006). Climatic or oceanographic conditions can have contrasting consequences for seabird species (e.g., Newton et al., 2009; Chambers et al., 2011; Krüger and Petry, 2011; Tavares et al., 2016). For example, upwelling processes, and resulting cold surface waters, may act differently on various prey species and thus may be beneficial to some seabirds (by increasing their prey availability) but increase mortality in other seabird species due to reduced food productivity (see Newton et al., 2009) or thermoregulatory stress. Understanding the impacts that climatic and oceanographic factors can have on seabird population decline is thus crucial but challenging.

Rainfall patterns are expected to change across the world as a result of climate change (Milly et al., 2005; Power et al., 2013). Droughts, in particular, are becoming more frequent and extreme in many regions of the world (Dai, 2013), resulting in reduced water outflow to the ocean and decreased oceanic productivity across multiple trophic levels (Lamberth et al., 2009; Lee et al., 2012; Arnell and Gosling, 2013; Auricht et al., 2018). Human interventions, such as water extraction and diversion, can greatly exacerbate climatic effects of drought, and even lead to hydrological droughts that are independent of climatic effects (Wada et al., 2013). Increased nutrient loading from freshwater outflows into the ocean stimulates phytoplankton and zooplankton production (Morgan et al., 2005), often leading to large aggregations of larval fishes and their predators near river estuaries (Cyrus and Blaber, 1987; Cyrus and Blaber, 1992; Grimes and Kingsford, 1996; Gillanders and Kingsford, 2002). Therefore, knowledge of the link between freshwater outflow and coastal trophic communities is critical but this relationship is often poorly understood (Gillanders and Kingsford, 2002; Gillson, 2011).

The Murray River is the largest river in Australia and a major source of freshwater entering coastal waters in South Australia. The Murray River discharges into two shallow and large freshwater lakes (Lakes Alexandrina and Albert also known as the “Lower Lakes”, regulated by a series of barrages), mixing in a shallow estuarine-lagoon (the Coorong) before reaching the mouth of the river (thereafter referred to as the “Murray Mouth”) ( Figure 1 ). Freshwater outflow into the coastal environment from the Murray River has declined significantly over the past century due to intensive drought periods and increased extraction for human use (Maheshwari et al., 1995; Thom et al., 2020). Human demands have reduced the average annual outflow at the Murray Mouth by 61%, meaning that water now ceases to flow to the mouth 40% of the time compared to 5% before increased extraction for human use occurred (Csiro, 2008). In addition, between 2001 and 2010, Australia experienced one of the worst droughts recorded since European settlement, known as the “Millennium Drought”. This combined pressure resulted in reduction of both seawater exchange and freshwater outflows (Mosley et al., 2012; Geddes et al., 2016), as well as decreasing coastal productivity in the area (Auricht et al., 2018).

Little penguins (Eudyptula minor) are colonial seabirds that become central-place foragers during breeding, with most of their prey being captured within less than 60 km of their colony when provisioning chicks (Collins et al., 1999; Bool et al., 2007; Hoskins et al., 2008). As a consequence, their breeding success is strongly correlated to food availability (Chiaradia and Nisbet, 2006) and large colonies are expected to be located near areas of high productivity, such as freshwater inputs or upwelling (see also Fortescue, 1999; Poupart et al., 2017). Tracking studies at several colonies in Victoria have shown that many little penguins have a tendency to forage near estuarine environments or freshwater outflows (Collins et al., 1999; Hoskins et al., 2008; Preston et al., 2008; Kowalczyk et al., 2015). Kowalczyk et al. (2015) in particular found that little penguins within Port Phillip Bay (Victoria) foraged closer to the Yarra River mouth during drought years, but increased their foraging range in years of high rainfall – when outflows from the Yarra River increased – likely to follow the dispersed nutrients (as nutrients play a crucial role in regulating primary production and hence the food web; Mortimer et al., 1999; Cruzado et al., 2002; Sommer et al., 2002) and thus maximize resource acquisition.

Little penguins are generalist feeders that rely mainly on clupeidae fish species, such as anchovy (Engraulis spp.) and pilchards (Sardinops sagax) (Klomp and Wooller, 1988; Cullen et al., 1992; Chiaradia et al., 2003; Chiaradia et al., 2012). In South Australia, anchovies are the most frequently consumed prey species and account for 31-67% of the little penguin estimated prey biomass (Bool et al., 2007; Wiebkin, 2012). Both anchovies (Wedderburn and Suitor, 2012) and little penguins (Colombelli-Négrel, unpublished data) have been recorded in the Coorong and Lower Lakes system of the Murray River, and the Murray Mouth is well within the foraging range of the little penguins breeding in Encounter Bay (20-60 km; average 41 ± 4 km; Figure 1 ; Bool et al., 2007). Therefore, it is likely that some dependence between the Encounter Bay penguins and productivity of the Murray River estuary exists. This hypothesis is further supported by the fact that penguin numbers in Encounter Bay have experienced major declines in the early 2000s (Colombelli-Négrel and Kleindorfer, 2014; Colombelli-Négrel, 2015a), which coincides with the Millennium Drought (2001-2010). While many factors have been suggested to explain these declines in Encounter Bay – such as predation on land or at sea, low reproductive success or juvenile survival, as well as parasite abundance (Bool et al., 2007; Wiebkin, 2011; Colombelli-Négrel, 2015b; Colombelli-Négrel, 2015a; Colombelli-Négrel, 2016) – only one study to date investigated the role of these suggested factors. Specifically, Colombelli-Négrel (2015a) reported that juvenile survival, but not reproductive success, was the most critical variable affecting little penguin population trends on Granite Island. Yet the relationship between penguin numbers in Encounter Bay and Murray River outflows has never been explored, which is the aim of this study. Based on the findings from Auricht et al. (2018), we hypothesised a significant and positive relationship between little penguin numbers and river outflow, via primary productivity and thus food availability (see conceptual model in Figure 2 ).

There was a clear reduction in daily river outflows from the Murray Mouth to the coastal ocean during the Millennium Drought, with minor discharges occurring between 2003 and 2006 and zero discharge between 2007 and 2010 ( Figure 3A ). High river outflows started to occur again following 2011-2012 ( Figure 3A ), which coincided with a stabilisation in the decline of little penguin numbers on Granite Island ( Figure 4 ). Chlorophyll-a concentrations in the ocean around the Murray Mouth and Granite Island were elevated in years with high river outflows and low in years when outflows were reduced ( Figures 3B, C ). A comparison of the chlorophyll-a concentrations in a year of extreme-drought period with no river outflow (low chlorophyll-a concentrations; 2006) vs a period of high river outflow (high chlorophyll-a concentrations; 2012) is presented in Figure 3C . Monthly sea surface temperature and daily rainfall temporal patterns are presented in Figures 3D, E respectively. There was a strong and distinct declining trend for both the annual and monthly little penguin numbers over time (monthly data z = -10.56, N = 192, P < 0.0001; annual data: z = -5.70, N = 38, P < 0.0001; Figures 4A, B ). For the monthly penguin numbers, while the data showed a small and significant increasing trend after 2010 (z = 7.65, N = 95, P < 0.0001; Figure 4A ), the overall declining trend (comparing before and after 2010) was stronger than the small increasing trend after 2010 (z = -8.56, N = 97, P < 0.0001; Figure 4A ).

The best model explaining the monthly penguin numbers included river outflow and sea surface temperatures ( Table 2A ). The model explained 39% of the variation in the data. Monthly penguin numbers were predicted to decrease by a multiplied factor of 0.94 for every 1 unit (°C) increase of sea surface temperatures (β = 0.94, 95% CI: 0.90; 0.97; Figure 5A ) and by a multiplied factor of 0.86 for every 1 unit (ML) increase of river outflow (β = 0.86, 95% CI: 0.84, 0.89; Figure 5B ). Analysing the direct relationships between the variables (see Figure 2 ), the best model explaining the decline of the annual penguin numbers included all predictor variables (river outflow, chlorophyll-a concentrations, rainfall, numbers of long-nosed fur seals and southern garfish commercial catch; Table 2B ). The model explained 74% of the variation in the data and the strongest predictors were the numbers of southern garfish and chlorophyll-a concentrations. Annual penguin numbers were predicted to increase in the model by a factor of 1.4 [= exp (0.33)] with increasing of 1 unit (tones) of southern garfish numbers (β = 0.33, 95% CI: 0.17, 0.50; Figure 6A ) but predicted to decrease by a factor of 0.37 [= exp (-1.01)] with increasing of 1 unit (mg/m³) of chlorophyll-a concentrations (β = -1.01, 95% CI: -1.65, -3.68; Figure 6B ). Annual penguin numbers were also predicted to increase with increasing river outflow (β < 0.001, 95% CI: 7.44e-07, 1.68e-05) and increasing rainfall (β = 0.01, 95% CI: 0.002, 0.03), but to decrease with increasing long-nosed fur seal numbers (β = -0.02, 95% CI: -0.003, -0.01). Adding the time lagged relationships described in Figure 2 did not change our results as the best model explaining the data remained model 3 ( Table 1 ). While the assumption of independence was satisfied in model 4 ( Table 1 ; with the time lagged relationships), the assumptions of constant variance and normality were not satisfied. In addition, the adjusted R² for model 4 was lower than for model 3 and the AIC and residual deviance of model 4 were higher than model 3 ( Table 2 ).

Seabirds around the world can serve as sentinels of environmental changes and help us understand how they are affecting wildlife (e.g., Wanless et al., 2007; Divoky et al., 2015; Thomsen and Green, 2019). In this study, we showed that (1) monthly little penguin numbers, an iconic and locally breeding Australian seabird, were negatively associated with sea surface temperatures and river outflow, and that (2) annual penguin numbers were positively associated with southern garfish numbers (our local indicator of food availability) but negatively associated with chlorophyll-a concentrations. The results of this study suggest that decisions regarding river water management should consider, not only human and terrestrial environmental requirements, but also the long-term impacts that this may have for the coastal environment outside the river system (see also Auricht et al., 2018; Hallett et al., 2018; Thom et al., 2020).

Chlorophyll-a concentrations are considered a reliable measure of primary productivity in an area (and hence food availability) because they are a proxy for phytoplankton production, which forms the base of the food chain as prey for zooplankton, fish larvae and other heterotrophic organisms (Monticelli et al., 2007; Lo-Yat et al., 2011; Lander et al., 2013). A previous study showed that decreasing Murray River outflow during the Millennium Drought had reduced primary productivity in adjacent coastal waters, and postulated there was likely consequences for higher trophic levels (Auricht et al., 2018). Specifically, the authors showed that primary productivity collapsed in the near-shore region during the Millennium Drought, and that the primary productivity plume was up to 60 km from the coast in years of high river outflows (Auricht et al., 2018), which would encompass Granite Island ( Figure 1 and 3B ). In our study, we found a negative relationship between chlorophyll-a concentrations and annual penguin numbers, which contrasts with a previous study on little penguins showing that annual survival increased with increasing chlorophyll-a concentrations (Agnew et al., 2015). As previously stated, chlorophyll-a concentrations have a near immediate response to nutrients from river outflow (Black et al., 2016), while there may be a time lag of up to 12 months between primary productivity and actual prey availability for marine top predators, such as seabirds (Price et al., 2020). Yet additional analyses including time lagged relationships (see Results as well as Tables 1 , 2 ) did not better explain the variation in penguin numbers. Interestingly, there was also a negative relationship between annual southern garfish numbers and chlorophyll-a concentrations. It is possible that other environmental variables, not accounted for in our study, may have impacted food availability (and hence this negative relationship) as small pelagic fishes (such as those preyed upon by little penguins) are often highly dependent on plankton-based food webs, which are very sensitive to short-term environmental changes in salinity (see Kelly et al., 2016). Furthermore, increased chlorophyll-a concentrations would increase the production of fish larvae, thereby also increasing the presence of predators and fishing intensity (see below) in the area, which in turn could have long-term negative impacts on little penguins. Finally, this negative relationship may be due to the low capacity of recovery of the Granite Island little penguin population as annual penguin numbers remained low and continued to decline after 2010 when chlorophyll-a concentrations started to increase. Indeed, previous study showed that the decline of the little penguins on Granite Island was linked to low juvenile survival (less than 2%; Colombelli-Négrel, 2015a). Delayed sexual maturity (in little penguins, individuals do not breed until they are two or three years old; Priddel et al., 2008) and a decrease in juvenile survival can add significant lagged effects on population size and recovery, especially for small population where birth and death rates can be extremely random (e.g., Legrendre et al., 1999). This suggests that the observed low juvenile survival may have had a stronger negative effect on little penguin numbers than any of the potential benefits gained from the increased river outflow (and consequent chlorophyll-a concentrations) once the population reached critical levels. Further investigation into the food web and association between South Australian little penguin numbers and chlorophyll-a concentrations are needed to fully understand their causal and temporal relationships.

Monthly little penguin numbers were negatively associated with sea surface temperatures and river outflow. This aligns with studies in South Australia (Johnson and Colombelli-Négrel, 2020) and Western Australia (Cannell et al., 2012) showing that little penguins had a lower breeding success when sea surface temperatures were higher in the months before breeding. However, studies in Victoria showed that foraging effort and first year survival in little penguins were positively associated with sea surface temperatures (Sidhu et al., 2012; Berlincourt and Arnould, 2015) and little penguins had a greater breeding success when sea surface temperatures were higher in the autumn preceding spring breeding (Cullen et al., 2009). The negative association with river outflow also contrasts with our annual results, as annual little penguin numbers were positively associated with river outflow. Monthly penguin numbers would have been affected by the overall annual decline of the population but also the probability that individuals be present on land at certain time of the year. Hence the contrasting results may be explained if greater numbers of penguins were fishing at sea for longer periods during times of high river outflow. Results similar to this were found by Kowalczyk et al. (2015) in Victoria where penguins increased their foraging range when the outflow from the Yarra River increased.

Seals are recognised predators of penguins worldwide (e.g., Hofmeyr and Bester, 1971; Du Toit et al., 2004; Johnson et al., 2006; Lee et al., 2019; Bester et al., 2020). In this study, we found that annual little penguin numbers decreased when long-nosed fur seal numbers increased, supporting the idea that seal predation may have influenced the decline of little penguins in South Australia (Bool et al., 2007; Wiebkin, 2011). Yet the relationship between penguin and long nosed fur seal numbers was close to 1:1, indicating that a decrease in penguin numbers due to seal predation would be very slow. While the available data may have underestimated the number of seals in the study area, previous studies have shown that the relative importance of little penguins in long-nosed fur seal diets varies significantly across regions (Bool et al., 2007; Baylis and Nichols, 2009; Reinhold, 2014), suggesting that little penguin availability is not the main factor driving predation. It is therefore possible that, in years of low river outflow when primary production was low and food resources may have been scarce, long-nosed fur seals switched to preying on little penguins (as seen in Antarctic fur seals (Arctocephalus gazella) and chinstrap penguins (Pygoscelis antarctica) when krill abundance was low; Daneri et al., 2008). An increase in long-nosed fur seal numbers may also have resulted in reduced food availability for little penguins as seals can compete with penguins for food resources (Croll and Tershy, 1998; Lowther et al., 2020) and long-nosed fur seals also consume southern garfishes in South Australia (Reinhold, 2014). These hypotheses should be further tested with additional tracking and diet studies.

Our model predicted an increase in annual penguin numbers with increasing southern garfish numbers, indicating that a decrease in food availability may have been responsible for the annual decline of the little penguin populations in the area (on Granite Island but also on Wright, West, Seal and Pullen islands; see Colombelli-Négrel and Kleindorfer, 2014). A previous study showed that juvenile survival was the most important driver of little penguin population trends on Granite Island (Colombelli-Négrel, 2015a). While no information is available for South Australian little penguins, studies in Victoria have suggested that mortality of juveniles in their first years occurred mostly at sea (Reilly and Cullen, 1982; Dann, 1991), likely due to a combined effect of parasites and starvation (Harrigan, 1992). These studies and our results suggest that low juvenile survival on Granite Island (less than 2%; Colombelli-Négrel, 2015a) may be driven by low food availability (see also Reilly and Cullen, 1982; Dann, 1988), which in turn would drive the observed populations declines. A study conducted in 2006 on Granite and West islands showed that garfishes were the second most frequent prey item consumed by little penguins (Bool et al., 2007). Yet, the foraging behaviour and diet of South Australian little penguins is little studied with only two studies to date (Bool et al., 2007; Wiebkin, 2012), and additional dietary studies would be beneficial, especially considering that southern garfishes are a commercially fished species in South Australia. Further research is therefore warranted to determine if commercial catch of southern garfishes may have negatively impacted little penguins.

In conclusion, we found a strong association between little penguin numbers and southern garfish numbers (our local indicator of food availability), which may explain the local decline observed in penguin numbers in Encounter Bay during the Millennium Drought. The fact that the Granite Island little penguin population still has not recovered in 2020 (Colombelli-Négrel, 2020), despite larger outflows in 2012-2013 and at the end of 2016, suggests that the population may have reached some critical reduction in the number of breeding birds during the drought period. While this study focused on a single colony, further decrease in little penguin numbers (as recorded in South Australia; Colombelli-Négrel and Kleindorfer, 2014; Johnson and Colombelli-Négrel, 2020, this study) may be observed in the future because of climate change and/or reduced rainfall/river outflow, considering their high dependence with estuarine environments and freshwater outflows (Collins et al., 1999; Hoskins et al., 2008; Preston et al., 2008; Kowalczyk et al., 2015). Much of the research to date on seabirds and climate change have focused on the impacts of increased ocean temperatures (e.g., Sandvik et al., 2005; Chambers et al., 2011; Descamps et al., 2017; Johnson and Colombelli-Négrel, 2020). Similar to other studies on Arctic systems showing the importance of glacial outflows on marine productivity as well as seabird foraging and populations (Grémillet et al., 2015; Urbanski et al., 2017; Bertrand et al., 2021; see also Michel et al., 2015), our findings highlight that other factors need to be taken into consideration, such as the potential importance of river outflow for the health and resilience of the coastal ecosystem, and our results should be considered in future river management strategies. Given the wide distribution of seabirds, their key role in ecosystems, and the fact that droughts are becoming more and more frequent (Dai, 2013), future studies, both within Australia and elsewhere, are needed to identify which species may be affected by hydrological droughts to further enhance both seabird conservation and river management.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Flinders University Ethics Committee (E388-E449).

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

We would like to thank the Adelaide and Mt Lofty Natural Resources Management Board, the Nature Conservation Society of South Australia, Birds SA, the Field Naturalists Society of South Australia, and the Lirabenda Endowment Research Fund for their financial support. The idea for this paper came from some work funded by the South Australian Department of Environment and Water (DEW) to Peter Dann.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.