Measuring the net biological impact of fisheries enhancement: pink salmon hatcheries can increase yield, but with apparent costs to wild populations

Despite increasing global demand for seafood, the production of marine capture fisheries has remained essentially stable over the past three decades (FAO 2014). Although there remains the potential for some increase by improved management (Worm and Branch 2012; Watson et al. 2013), the plateau in capture fisheries is generally believed to reflect a fundamental limitation in the capacity of the world’s oceans to generate food (Worm et al. 2009; Chassot et al. 2010; Worm and Branch 2012). Intensive aquaculture appears to offer an opportunity to circumvent this limitation, and indeed, the rapid expansion of the industry has allowed continued growth in fish production in recent decades (FAO 2014). However, in many cases intensive culture of marine species is infeasible for technical, economic, or political reasons (Bostock et al. 2010). The enhancement of wild populations through release of hatchery-reared juveniles is an intermediate approach that has been practiced in a variety of marine fish and invertebrate species for over a century (Hilborn 1998; Bell et al. 2006; Lorenzen et al. 2013). Also known as stock enhancement or ocean ranching, this type of aquaculture generally involves the rearing of juveniles in a hatchery past some critical stage before release in to the wild, thereby circumventing high levels of mortality or habitat limitations associated with early life-history stages (Leber et al. 2004). Surviving individuals are then expected to be available for capture after several years of ocean growth. These methods may also be suitable for hastening the recovery from historical overfishing (Molony et al. 2003). As such, there is a great deal of interest in the use of fisheries enhancement to rebuild depleted fisheries and to bolster the productivity of healthy stocks. However, despite a long history of experimentation, successful enhancement of marine species is rare, and most efforts remain in a research and development phase (Lorenzen et al. 2013; Trushenski et al. 2014).

In contrast with the limited success of marine stock enhancement, large-scale hatchery programs for anadromous salmonids — especially Pacific salmon (Oncorhynchus spp.) — have been operating for decades, and today it is estimated that nearly one in four salmon in the Pacific Ocean are of hatchery origin (Larkin 1974; Ruggerone et al. 2010) and overall abundance of Pacific salmon in the ocean has increased greatly (Wertheimer et al. 2005; Ruggerone et al. 2010; Peterman et al. 2012). Despite the long history and massive scale of hatchery salmon production, the efficacy of salmon enhancement programs as a tool for increasing fisheries productivity has rarely been rigorously demonstrated (Larkin 1974; Lorenzen 2005; Naish et al. 2007; Paquet et al. 2011). It has been repeatedly suggested over the past 30 years that to improve enhancement efforts it is necessary to specify clear, measureable goals and monitor outcomes relative to these goals (Peterman 1991; Hilborn 1998; Naish et al. 2007; Paquet et al. 2011). Nevertheless, monitoring and evaluation of salmon hatchery programs remains largely insufficient (Naish et al. 2007). To ensure that expected enhancement effects are being achieved, evaluation of hatchery programs must consider all relevant risks and benefits (Hilborn 1998).

Much of the difficulty in evaluating salmon hatchery programs results from a lack of suitable controls that would allow for isolation of any enhancement effect. Manipulation of stocking rates provides one avenue for distinguishing environmental and hatchery influences on fisheries production (Buhle et al. 2009), but experimental reductions in hatchery production are typically precluded by legal, political, or economic considerations (Naish et al. 2007). Retrospective analyses that attempt to explain trends in abundance using time series of environmental variables and stocking rates have become more common as data on enhanced populations is increasingly available (Wertheimer et al. 2004; Morita et al. 2006; Scheuerell et al. 2015). Alaska’s pink salmon (Oncorhynchus gorbuscha) enhancement programs provide a unique opportunity to examine the net biological impact of large-scale stock enhancement both because of its scale and the quality of available data. Compared with other salmon-producing regions in the Pacific, Alaska’s hatchery programs are relatively young, and as a result reliable catch and abundance data exist for both pre- and posthatchery periods (Olsen 1993). Additionally, since 1995 most hatchery pink salmon have been thermally marked, which allows for reliable attribution in the catch (Hilborn and Eggers 2000). Hatchery releases began during the mid-1970s (Olsen 1993) and combined releases from the two largest programs have been stable around 750 million since about 1990 (Brenner et al. 2012). These programs constitute around 10% of the total number of salmon juveniles released to the North Pacific and more than half the total pink salmon (NPAFC 2016; Fig. 1; Table 1).

Four regions account for the majority of the pink salmon catch in Alaska. Prior to hatchery supplementation, pink salmon were most abundant in Southeast Alaska (SEAK; ∼20 million annual run), followed by Kodiak (KOD; ∼10 million), Prince William Sound (PWS; ∼7 million), and the south Alaska Peninsula (SPEN; ∼3 million). Enhancement occurs in three of the regions, though the scale of operations varies by orders of magnitude. PWS pink salmon is currently the largest hatchery program in the world by annual number of releases (NPAFC 2016). Since 1990, on average, 77 (SD = 48) hatchery fry have been released for each returning wild adult fish in PWS, while in KOD this ratio is about 8:1 (SD = 3.6) and in SEAK close to 1:1 (SD = 0.5). Unlike many other regions where hatcheries are intended to mitigate declines in salmon populations resulting from habitat degradation, Alaska’s hatcheries are designed to produce harvestable fish to supplement relatively healthy wild populations (Naish et al. 2007). Since the inception of the hatchery programs, pink salmon catches have increased dramatically, especially in PWS where hatchery returns now average over 35 million fish and peaked at 76 million in 2013. The majority of these fish are harvested in common-property commercial fisheries, though hatchery operators also harvest on average 30% of returning fish to cover production costs (Botz et al. 2013). Despite the ostensible success of enhancement, uncertainties regarding impacts of hatchery-origin fish on wild salmon and other species continue to cause concern among many stakeholders (Pearson et al. 2012; Brenner et al. 2012; Jasper et al. 2013). Since 2012 these concerns have contributed to delays in the recertification of Alaska salmon by the Marine Stewardship Council, resulted in a “Category C” grade for PWS salmon from the Fisheries Sustainability Partnership, and motivated an intensive research program by the Alaska Department of Fish and Game.

Recent analysis of hatchery programs from around the Pacific have found limited evidence of a large enhancement effect and in many cases identified concerns about negative impacts on wild populations. For example, Morita et al. (2006) modeled pink salmon catch in relation to hatchery output and climate factors and found that intensive stocking contributed little to a dramatic increase in abundance after 1990. Ohnuki et al. (2015) used tagging data to confirm the minor contribution hatchery-origin fish to commercial pink salmon catches in Japan and suggest that the costs of hatchery production likely outweigh the benefits. Kaev (2012) examined the population dynamics of chum (Oncorhynchus keta) and pink salmon in the Sakhalin–Kuril region of Russia and found evidence of an enhancement effect in hatchery-supplemented chum populations, but not in pink salmon populations. Sahashi et al. (2015) found that hatchery stocking of masu salmon (Oncorhynchus masou) in the Shari River tended to displace rather than supplement natural production. Similarly, Scheuerell et al. (2015) compared supplemented and natural populations of Snake River Chinook salmon (Oncorhynchus tshawytscha) and identified only minor increases (∼3% on average) in adult density attributable to enhancement efforts. Buhle et al. (2009) identified negative impacts of hatchery coho (Oncorhynchus kisutch) on wild Oregon coast populations and documented increased wild productivity following large reductions in hatchery supplementation. Finally, Zhivotovsky et al. (2012) used genetic and demographic analyses to show that rapid expansion of a chum hatchery program on Iturup Island led to the extirpation of a distinct beach-spawning ecotype by abundant hatchery strays.

Given the limited success demonstrated by these recent hatchery studies, it is not surprising that the net biological impact of Alaska’s pink salmon hatchery programs have been a matter of considerable debate. Consistent with reports of limited benefits of hatchery programs, several previous studies have concluded that improved ocean survival associated with a large-scale shift in marine environmental conditions would have led to increased pink salmon catch even in the absence of hatchery production (Eggers et al. 1991; Tarbox and Bendock 1996; Hilborn and Eggers 2000, 2001). Others have argued that hatchery production is primarily responsible for increasing catches and conclude that the enhancement program is highly successful (Smoker and Linley 1997; Wertheimer et al. 2001; Heard 2003; Wertheimer et al. 2004). Hilborn and Eggers (2001) describe these two hypotheses as “augmentation” and “replacement”. Under the augmentation hypothesis, hatchery production adds additional productivity to the fishery without impacting existing wild stocks. Alternatively, the replacement hypothesis asserts that hatchery production reduces wild stock productivity, and thus hatchery fish effectively replace wild fish in the catch. In practice these hypotheses define the ends of a gradient; under complete replacement the net value of one hatchery fish approaches zero, while under complete augmentation each hatchery fish could be considered equal to one additional wild fish.

To make predictions about the trajectory of the PWS pink salmon fishery in the absence of the hatchery program, we rely on two key patterns of productivity in salmon populations. First, oscillation between North Pacific climate regimes has been shown to predictably influence salmon abundance (Hare et al. 1999; Beamish et al. 1999, 2004). Second, covariation in the productivity of salmon stocks has been shown to be highest in geographically proximate populations (Pyper et al. 2001; Wertheimer et al. 2001), and indeed Alaskan salmon populations have shown strong spatial coherence in decadal-scale patterns of productivity (Hare et al. 1999). Thus, to establish an empirical estimate of net biological benefit, we examine over 50 years of catch and abundance data from four pink salmon-producing regions in Alaska (Fig. 2) to predict catch and wild stock productivity in the absence of enhancement efforts. The present study builds on previous reviews of Alaska’s pink salmon hatcheries, including Eggers et al. (1991), Hilborn and Eggers (2000), and Wertheimer et al. (2001), and benefits from over 15 recent years of data, a period of consistently intense hatchery stocking (Fig. 1). With this extended data set, we are also able to compare spawner–recruit relationships for wild pink salmon populations before and after the implementation of hatchery programs. In addition, we consider the impact of hatchery production on interannual variability in pink salmon abundance.

Enhancement of pink salmon in Alaska — particularly in the PWS management area — has succeeded in producing a substantial and sustained enhancement effect and contributed to an order of magnitude increase in catch since the 1960s. At the same time, local wild populations have remained “sustainable” insofar as their abundances remain stable and they appear at no immediate risk of collapse. While increased variability in catch resulting from high abundances may be problematic from a fisheries and processing perspective, overall the hatchery program appears to provide a net contribution to harvest. However, our results also demonstrate that if reduced wild productivity and the costs of hatchery production are not accounted for, the benefits of enhancement may be considerably overestimated. The magnitude of increased catch in PWS has been at least twice as great as nearby areas, implying a large contribution from hatchery production, but comparisons with adjacent regions also suggest that favorable ocean conditions would have resulted in an increasing abundance trend even in the absence of an enhancement program. Therefore, in the case of PWS, although the mean catch of hatchery fish since 1990 has been 30 million, our best estimate of the net enhancement effect to the commercial fishery (9 million) is less than one-third of the apparent contribution when impacts on wild production and cost-recovery are ignored. In regions with smaller hatchery programs — KOD and SEAK — our models suggest a negligible contribution of hatcheries to increased catches. Thus, overall our results are consistent with previous studies that find enhancement effects of salmon hatcheries to be relatively minor (Morita et al. 2006; Scheuerell et al. 2015) and context-dependent (Kaev 2012).

The utilization of adjacent management areas as pseudo-replicates leaves the possibility that some local phenomenon has caused the atypical trajectory of wild pink salmon productivity in PWS and KOD. At a larger spatial scale, wild pink salmon populations from throughout the species range have increased in abundance by an average of 90% since the 1976–1977 ocean regime shift, further suggesting some unique factor at play in PWS and KOD (Morita et al. 2006; Ruggerone et al. 2010). A continued upward trend in hatchery returns despite relatively steady release levels since 1990 demonstrate that local marine conditions are not limiting productivity in hatchery pink salmon. Some persistent change in the productivity of the freshwater life-history phase would therefore be required to explain constant productivity despite improved marine conditions. Wertheimer et al. (2001) posited that the 1989 Exxon Valdez oil spill could account for the divergent pattern of abundance in PWS wild pink salmon. However, recent estimates of the impact of the spill on PWS pink salmon are modest, and the populations have been considered fully recovered from spill impacts since 2002 (Quinn et al. 2002; Brannon et al. 2012; EVOSTC 2014).

Based on our analysis of wild pink salmon productivity in Alaska, we conclude that the release of hatchery pink salmon has likely reduced productivity of the wild populations that interact substantially with hatchery salmon. While wild stocks in the SPEN and SEAK regions experienced dramatic increases in MSY (∼200%) — apparently as a result of increased carrying capacity (Fig. 5) — no such increases were observed in PWS or KOD. This pattern suggests that natural carrying capacity may have also increased in PWS and KOD, but is utilized by hatchery fish and thus no change is apparent for the wild stocks, essentially the pattern predicted by the replacement hypothesis (Hilborn and Eggers 2001). Our analyses do not, however, implicate any particular mechanism for negative impacts of hatchery–wild interaction. Understanding the mechanism or mechanisms by which hatchery production reduces wild stock productivity is critical for quantifying the long-term risk to wild stocks and identifying appropriate management responses. If reduced productivity is primarily a result of ecological interactions that reduce wild pink salmon survival or spawning success, then wild stocks would presumably recover quickly in response to reduced hatchery releases. Although salmon are well known for their ability to reliably return to their natal streams, some proportion of a population will enter and spawn in other streams, a phenomenon known as straying (Westley et al. 2013). Hatchery salmon commonly stray and often interbreed with wild conspecifics, but generally produce fewer successful offspring than their wild counterparts (Naish et al. 2007; Christie et al. 2014). The long-term effects of regular hatchery introgression are uncertain, but in any case genetic impacts on productivity would be expected to persist for multiple generations (Grant 2011; Baskett and Waples 2013; Harbicht et al. 2014).

Previous studies have identified the potential for both ecological and genetic interaction between hatchery and wild pink salmon in Alaska. As noted previously, despite widespread marking of hatchery pink salmon in Alaska, there is no systematic effort to quantify rates of straying by hatchery fish. However, recent studies have provided evidence that straying rates by PWS hatchery pink salmon may be significant. Brenner et al. (2012) found that in some PWS streams up to 98% of fish on spawning grounds were of hatchery origin. The degree to which these fish successfully breed with wild individuals is unknown, but recent genetic analyses have found significant hatchery introgression in PWS and SEAK wild chum salmon (Jasper et al. 2013). Notwithstanding breeding success, high rates of straying reduce the validity of escapement estimates and can therefore diminish the effectiveness of wild stock management.

High straying rates indicate large potential for ecological or genetic interaction between hatchery and wild fish and also confound efforts to estimate wild escapement. With hundreds of millions of hatchery releases occurring in PWS, even low absolute straying rates can result in high proportions of hatchery fish on some wild spawning grounds. Further research on the prevalence of straying and the genetic contribution of hatchery strays to the wild gene pool should be a priority. Though high stray rates implicate reproductive interaction as a likely mechanism for hatchery impacts, interactions at other life-history stages should not be ignored. The period immediately following ocean entry is thought to be very important to lifetime survival of anadromous salmon, and localized resource depletion by large numbers of hatchery fry may potentially impact growth and survival of wild fish (Cross et al. 2008). Increased abundance is also thought to be driving a downward trend in adult body size in PWS hatchery and wild pink salmon, which suggests competition during ocean rearing and homeward migration (Wertheimer et al. 2005). Taken together, these various interactions between hatchery and wild pink salmon demonstrate that a variety of plausible mechanisms exist for hatchery program impacts on wild productivity. It seems feasible that with improved understanding of these mechanisms, an effective accounting of the benefits and risks of hatchery operations for Alaska salmon enhancement could be accomplished. However, because pink salmon migrate long distances and potentially interact with many other salmon populations and species, the net benefits of enhancement will ultimately be sensitive to the geographic scope of analysis.

There is accumulating evidence that pink salmon have far-reaching impacts on ocean ecosystems. Patterns of alternating abundance in species that share ocean habitat with pink salmon strongly suggest impacts of competition (Ruggerone and Nielsen 2004). Such patterns have been observed in other salmon species, including comparatively valuable Bristol Bay sockeye salmon (Oncorhynchus nerka) (Ruggerone et al. 2003) and threatened Puget Sound Chinook salmon populations (Ruggerone and Goetz 2004). Recent analysis of long-term data on seabird populations in the North Pacific demonstrated similar patterns in reproductive success, implying that pink salmon also compete directly or indirectly with higher trophic levels (Springer and van Vliet 2014). There is also growing concern that large hatchery releases from around the North Pacific may be resulting in density-dependent declines in growth and survival for all salmon species as oceanic carrying capacity is approached (Cooney and Brodeur 1998; Kaeriyama et al. 2009). When considered in this broader ecosystem context, the analysis of stock enhancement becomes much more complex. With an increasing focus on ecosystem-based management of the oceans, the broader impacts of future enhancement efforts are likely to be heavily scrutinized (Pikitch et al. 2004; Samhouri et al. 2014). Ultimately, if these efforts are to be compatible with ecosystem-based principles, it will be critical to understand the biological capacity for enhancement and the potential unintended consequences of large-scale hatchery releases.