Following the review of this project (ISRP 2018), the ISRP suggested that the ‘program should not remain static’ and that we could better address emerging priorities through modifications of the study design and improved survey methods. We acknowledge, and agree with, their conclusion and have spent considerable effort to update and improve our overall study design. Our goal for this section of the proposal is to describe our vision for an integrated ecosystem study of marine juvenile salmon. Below, we provide a description of the general framework within which we propose to work, followed by a set of three funding scenarios (Options A, B, and C) describing the potential work that could be accomplished under different budget amounts. Please note that the Objectives and Budget listed in other sections of this proposal do not allow for multiple funding Options, and therefore represent Option B from the funding options (see “Options for future work” and Table 2 below).

Within this vision, we will describe current on-going work, work that is at risk due to budget cuts/cost increases, and new objectives (including some work that was previously funded, but cut due to past budget constraints as well as priorities that have emerged over the past several years). We do not expect BPA to be responsible for covering the full cost of this integrated ecosystem study. Instead, we explicitly describe project objectives that are currently supported by BPA and suggest how BPA might support these objectives going forward. We also describe what objectives are either in part or fully covered by NMFS cost share. Finally, we will describe what pieces of this integrated study are not currently covered, for which we don’t have clear funding avenues identified. We seek advice and comment from ISRP and NWPPC to identify potential sources of financial support and collaboration. We have undertaken this full-scale description of potential juvenile salmon marine ecology studies as we agree with the current ISRP review of our work and aim to address some of the suggestions for improvements related to ecosystem-level processes identified in that review.

Technical and/or scientific background

Evidence continues to accumulate suggesting that growth and survival of salmon during their early ocean life history phase is dependent upon ocean conditions (Pearcy, 1992, Kareiva et al., 2000, Peterson & Schwing, 2003, Beamish et al., 2004, Mueter et al., 2005, Pyper et al., 2005, Burla et al., 2010a, Burke et al., 2013b, Miller et al., 2014, Daly et al., 2017) and resulting forage availability (Dale et al., 2017). Several events over the last 15 years have highlighted the central role of the ocean environment for salmon population dynamics. In 2007, the collapse of the Sacramento River stocks due to poor ocean conditions resulted in the closure of all ocean fishing in 2008-2009. Meanwhile, we saw near-record returns of Chinook and coho salmon to the Columbia River in 2009 and 2010. More recently (2013-2016), a marine heat wave in the North Pacific Ocean (termed ‘the blob’; Bond et al., 2015) resulted in a complete shift in the pelagic community off Oregon and Washington, also resulting in severe declines in returning salmon abundance. Due to extensive media coverage concerning climate variability and salmon survival in relation to these events, stakeholders and the general public are now more aware of the role that varying ocean conditions can play in the recovery of listed salmon stocks. Recent research, largely by our team, points to ‘ocean conditions’ as the key factor affecting return rates.

An Ecosystem Framework

As the California Current continues to warm, acidify, and deoxygenate with climate change, we can also expect changes in ecosystem processes and components. These include shifts in the phenology, distribution, and productivity of salmon and their predators, forage, and fisheries (Auth et al., 2018). In light of these emerging and expected changes, a more holistic, ecosystem approach to understanding ocean impacts on salmon is vital for successful management and recovery. In general, NOAA Fisheries is currently charged with implementing an ecosystem-based approach to fisheries management that better integrates biological, physical, and social factors in assessments of fish stocks (NOAA Fisheries, Annual Guidance Memo 2017).

To date, most knowledge of relationships between ocean conditions and salmon survival relies on correlations, which have proven to break down over time (due either to changes in the ecosystem not represented by the correlation or usage of a correlation that was spurious to begin with). To advance our knowledge, mechanistic models of the ecological processes acting in the marine environment need to be developed. The ecological framework we propose to work under is graphically described in Figure 1 (taken from Wells et al., In prep.). This conceptual approach can be used to prioritize research and monitoring activities that are directly relevant to management decisions.

Figure 1. Conceptual Diagram of the role of ecosystem components (colored boxes) in the salmon life cycle, with a focus on the marine life stage. Processes that can be described with research or monitoring are shown as arrows connecting the ecosystem components, each one with a host of associated hypotheses about mechanistic relationships (see Table 1). Also shown are how each component relates to management drivers of potential change (grey boxes). Figure 1 and Table 1 were taken from Wells et al. (In prep.).

This framework explicitly links the multiple habitats that salmon use throughout their life. Thus, it is ideal for addressing the connections among these habitats and concomitant carry-over effects. For example, we developed a set of Research Foci in Table 1 that address how freshwater practices that influence smolt size or outmigration timing can have an effect on subsequent marine survival (Focus 1), and address how management of avian, piscivorous, and mammal predators will have a direct impact on salmon survival (Foci 8 and 9). This is particularly relevant with the recent efforts by Washington State’s Orca Task Force to simultaneously manage orca, other marine mammals, and Columbia River Chinook populations.

Importantly, many of the Research Foci identified in Figure 1 have not been extensively studied. It is for this reason that we (everyone in the Columbia River Basin) rely heavily on correlation analyses to estimate potential adult salmon returns. Although currently helpful, the usefulness of these correlations will degrade over time, particularly with the anticipated changes in ocean processes expected under climate change.

Our approach will be to use ecosystem models, parameterized to the extent possible for Columbia River salmon stocks, to estimate a range of possible biological and behavioral shifts in response to environmental perturbations and management practices. For example, if salmon stocks alter outmigration timing due to altered flow or temperature profiles in the Columbia River, we can use the data and results we have in hand to estimate the impact of an early or late outmigration on ocean growth and survival. With this approach, we can estimate best and worst-case scenarios, giving managers better information for planning and decision making. Importantly, we can also highlight potential biological responses to which salmon are particularly sensitive as well as those that do not have a large impact on salmon. For example, salmon survival may be particularly sensitive to the spatial distribution of a particular predator or the timing of arrival for particular prey species. Once identified, these can be better monitored and become part of an early warning system for management decisions, as well as direct future research and monitoring efforts.

The need for an ecosystem-based approach to fisheries management (EBFM) that incorporates ecosystem considerations, including species interactions and oceanographic conditions, is widely accepted within NOAA Fisheries. In the California Current Ecosystem (CCE), several models have been developed to serve as decision support tools for EBFM (either in isolation or a multi-model framework). For example, food web models (e.g. Ecopath) are being used to identify critical ecological interactions of target and non-target species in the CCE and to inform management about potential trade-offs among fisheries (Field et al., 2006, Ruzicka et al., 2012, 2016, Kaplan et al., 2013, Koehn et al., 2016). Whole ecosystem models (e.g. Atlantis) are being used to examine hypotheses about how fish populations and the broader ecosystem respond to different management and climate scenarios (Marshall et al., 2014, 2017, Kaplan et al., 2017, Hodgson et al., 2018). In addition, Models of Intermediate Complexity for Ecosystem assessment (MICE), which bridge the gap between full ecosystem models and traditional single species models, and are being used to evaluate ecosystem and fishery consequences of alternative management strategies (Plagányi et al., 2014, Punt et al., 2016) and represent a good option for this project to focus initial modeling efforts.

Ecosystem models are an evolving discipline in fisheries management. Until recently, many fisheries assessment models used around the world have ignored or made overly simplistic assumptions about trends in predation and other ecosystem processes that affect the productivity of commercially, recreationally, or culturally important fish stocks (Skern-Mauritzen et al., 2016). However, evidence of for predation affecting harvested fish stocks continue to expand (e.g., consumption of herring (Clupea harengus) and forage fish by large whales in the northeastern U.S. (Overholtz & Link, 2006) and consumption of salmon (Oncorhynchus spp.) and forage fishes by pinnipeds in the northwestern U.S. (Wright et al., 2007, Thomas et al., 2011).

Estimating ecosystem interactions, and predation in particular, improves our understanding of predator - prey relationships and can inform fishery management reference points (Hollowed et al., 2000, Tyrrell et al., 2011). The goal of our future research is to focus on several types of predation, including marine mammals, birds, and fishes. Research has shown that Chinook salmon recovery may be limited by predator aggregations: avian predators along the Columbia River are thought to consume 5-12 million Chinook salmon juveniles annually (Roby et al., 2003), spiny dogfish that congregate near hatcheries in British Columbia are thought to consume between 0.5 - 7 million juvenile salmon annually (Beamish et al., 1992), and marine mammals have a clear and increasing impact on both juvenile and adult salmon (Chasco et al., 2017a, Chasco et al., 2017b). Additional Chinook salmon predators include dogfish, salmon sharks (Nagasawa, 1998), and the marine birds, common murre and sooty shearwater (Phillips et al., 2018). Thus, quantifying the magnitude of marine predation and putting this in the context of other factors impacting salmon and steelhead is therefore increasingly important.

With this in mind, we created four Objectives for the current proposal, each tied to multiple Research Foci in our ecological framework (Figure 1, Table 1). Again, this framework provides the tools to prioritize research and monitoring efforts towards a goal of better understanding how ecosystem processes impact salmon and how management actions my influence those impacts. The four Objectives of our proposal are:

Although these Objectives address the majority of the Research Foci, it is important to note that the extent to which each Foci is addressed depends on decisions made regarding funding. Here, we describe many of the Ecosystem Components depicted in Figure 1, discussing existing knowledge as well as how we might advance our understanding of them with an ecosystem approach. We then describe 3 options for research and monitoring rooted in our ecosystem framework, each with an associated budget (see Options A, B, and C below). To the extent possible, we link the activities within each option to historical context (current work, current work at risk due to budget cuts/cost increases, and new proposed work).

Physical Environment

Physical ocean conditions include factors at multiple scales, such as ocean basin-scale winds and ocean circulation patterns, regional scale water temperature and upwelling, and mesoscale eddies and fronts. Each of these factors varies in linear and non-linear ways at seasonal, interannual, and decadal time scales.

At the basin scale, the North Pacific Ocean has historically experienced dramatic shifts in climate and productivity at a frequency of 20-30 years, caused by changes in wind speed and direction that are partially related to the position and intensity of the Aleutian Low Pressure system in winter. These phase shifts are indexed to some extent by the Pacific Decadal Oscillation. During the cool regime that extended from the 1947-1976, salmonid stocks were low in the subarctic, but high in the California Current (Francis & Hare, 1994, Francis et al., 1998). The presence of a warm phase of the PDO throughout the 1980s and most of the 1990s may well explain the very low survival of Oregon Production Index coho (all coho populations south of Leadbetter Pt, WA) during that period and poor returns of spring Chinook to the Columbia River. Similarly, the recent anomalous warming (the ‘blob’) was associated with extended positive values of the PDO; we are still documenting the negative impact of this marine heat wave on Columbia River salmon stocks (see Daly et al., 2017, Morgan et al., In press).

Another example of extreme disruptive processes are El Niño events, the last of which (2015/2016) was one of the largest on record. The consequences of El Niño events are well known: reductions in biological productivity followed by high mortality of pelagic fishes, salmon, and seabirds. The recent variability in basin-scale ocean conditions (PDO, El Niño, the ‘blob’) since 1998 has provided us with a natural experiment in which we can compare salmon growth and survival to the physical, biological and ecological changes in the ocean that might impact salmon growth and survival. Although we observed record low Catch per Unit Effort (CPUE) of yearling coho and Chinook salmon in our 2017 sampling, we cannot yet identify specific mortality agents. The best way to take advantage of these natural experiments is to place these observations into our ecosystem framework, accounting for all potential drivers of salmon survival simultaneously.

At the other end of the spectrum (meso-scale), the Columbia River plume may affect growth and survival of juvenile salmonids in multiple ways. For example, the shape and extent of the plume is controlled by the amount of freshwater flowing out of the Columbia River, by local wind and by ocean currents. The plume is typically oriented northward along the Washington coast during late fall and winter, and southward and offshore of the Oregon coast during summer, although changes in upwelling winds causes high seasonal variability (Burla et al., 2010b). Adult returns of fall Chinook salmon are positively correlated to greater plume volumes (Miller et al., 2013). The presence of the plume and the fronts associated with it tend to aggregate large numbers of forage species and predators, such as sooty shearwaters and common murres (Zamon et al., 2014, Phillips et al., 2017). The piscivorous seabirds also track changes in the shape and extent of the plume (Phillips et al., 2018). Ultimately, these ecological dynamics impact salmon movement and survival (De Robertis et al., 2003, 2005, Miller et al., 2013, Brosnan et al., 2014).

Experiences of salmon in freshwater may affect their survival in the ocean (Gosselin et al., 2018). These carryover effects are especially relevant for understanding how passage through a hydropower system results in altered marine survival for fish migrating in the river versus those transported by barge. Similarly, effects of Columbia River temperature and flow on salmon movement and growth can carry over to influence salmon spatial and temporal matches with optimal marine prey resources (e.g., spring Chinook salmon migrated out of the Columbia River almost two weeks earlier than normal in 2016, resulting in anomalous coastal spatial distributions). Finally, habitat conservation and restoration, particularly in the Columbia River estuary, may give salmon a boost in growth that can have repercussions for predator avoidance in the marine environment.

In future field sampling and data analyses, we propose to address some of the remaining questions related to the physical environment, including aspects of the freshwater and estuarine environments over which we have some control. See Research Foci 1, 2, and 3 in Table 1 for specific examples.

Primary Production

A productive and diverse ecosystem relies on nutrients and organic matter ultimately drawn from primary producers. In the NCC, most primary production occurs in the spring and summer, resulting largely from wind-driven upwelling, which brings nutrient-rich water to the surface. As these nutrients are taken up by phytoplankton and work their way up the food web, salmon and other pelagic organisms benefit in the form of increased food supply and growth potential.

Ecosystem processes related to bottom-up drivers are critical for salmon growth, resulting in interannual fluctuations in growth rate and mean size (Beamish et al., 1992, Tomaro et al., 2012, Miller et al., 2014, 2017, Daly & Brodeur, 2015, Daly et al., 2017, Litz et al., 2018). Importantly, these dynamics ultimately affect salmon survival, resulting in a correlation between measures of primary production (often in the form of chlorophyll concentration) and salmon distribution (Bi et al., 2007, Beacham et al., 2008, Bi et al., 2008, Yu et al., 2012, Burke et al., 2013b) and survival (Burke et al., 2013a). Salmon don’t rely on chlorophyll directly; there are multiple trophic interactions that must occur (zooplankton, larval and juvenile fishes, etc.), each of which varies from year to year. Chlorophyll concentration is used merely as an indicator of the potential for the ecosystem to produce biomass at higher trophic levels (Ware & Thomson, 2005). Therefore, we consider these base ecosystem components to be necessary for a productive system, but not sufficient by themselves to result in high salmon survival. Nor are they sufficient by themselves as indicators of ocean conditions for salmon management. In an ecosystem framework, these data are informative, but must be linked mechanistically to other ecosystem components.

We propose to continue collecting these important data streams going forward. In addition to their importance in our time series approach to monitoring the ocean ecosystem, there are multiple aspects of how primary production impacts the rest of the system, and salmon specifically (see Focus 4 in Table 1 for examples).

Forage Species

For purposes of this proposal, we define forage species in a salmon-specific manner, incorporating nektonic organisms such as krill, crab megalopae, and market squid as well as larger prey items such juvenile anchovy, rockfish, and smelt – all organisms that salmon smolts prey upon (Daly et al., 2009). Much like primary productivity, metrics summarizing presence and availability of these organisms can be used as indicators of salmon growth potential and can inform management decisions (see stoplight chart: https://www.nwfsc.noaa.gov/research/divisions/fe/estuarine/oeip/g-forecast.cfm#TableSF-02).

Yet, understanding the role of forage for salmon dynamics, particularly in a changing ocean environment, requires more detailed information about abundance, spatial distribution, composition, and ultimately the ecological drivers of forage organisms. Differences between the diets of juvenile salmon from along the California Current are attributable to the variability in forage dynamics (Hertz et al., 2015), which relate to environmental conditions prior to and during salmon out-migration (Daly et al., 2013, Friedman et al., 2018).

An important aspect of trophic dynamics is the phenology of forage availability. The majority of the prey that salmon consume are winter spawned fish and crab larvae. Winter ocean conditions prior to salmon outmigration are key to the development of salmon forage (Daly et al., 2013). Near the Columbia River, the spring transition marks the beginning of the productive, upwelling wind-driven spring and summer seasons. Abundance and quality of salmon forage are part of this phenology, resulting in periods of optimal and suboptimal forage availability. Over many generations, salmon have adapted to migrate from the Columbia River, arriving in the ocean as these prey dynamics peak, or in time to match the peak prey availability farther north. Understanding the mechanisms behind these relationships, and in particular, how they might vary with climate change, is vital to estimating future trends in salmon growth and survival.

Sampling the full suite of forage species over space and time is beyond the scope of any single sampling effort. This would require multiple types of sampling (e.g., sampling salmon and prey concurrently in space and time, acoustic surveys over larger spatial scales, interannual summaries of forage fish population dynamics, etc.). This could be addressed by integrating results from existing efforts (e.g., Coastal Pelagic Species survey, SWFSC; Pre-recruit survey, NWFSC) with increased sampling from the JSOES project. In addition, an increased use of the NOAA vessel RV Shimada, which has integrated acoustic gear, (Simrad EK60/EK80 multifrequency echosounders with transducers mounted on a retractable centerboard and operating at 38, 70, 120, and 200kHz), could assist in creating a spatially explicit description of krill and other pelagic prey organisms. Expanded use of the RV Shimada is a major focus for NMFS cost-sharing opportunities in near future.

Alternative prey for salmon predators. The life history of several forage fish species, such as northern anchovy, is such that some individuals will be salmon prey, some will be competitors of salmon, and others will be alternative prey for predators of salmon, complicating the ecological relationship between salmon and forage fish (Emmett et al., 2006, Daly et al., 2009, Hill et al., 2014). Recent evidence suggests the role of forage fish can mediate the impact of top predators (Emmett & Sampson, 2007, Wells et al., 2017) with significant impacts on salmon mortality rates. Given the multiple Ecosystem Components into which forage fish can be placed, any ecosystem approach to salmon management should include research and monitoring of this important guild of fishes. Indeed, the recent ISRP review of the JSOES project suggested additional research into forage fish as alternative prey (ISRP 2018).

Competitors

Interspecific competition has been shown to have a large effect on salmon growth and survival (Ruggerone & Nielsen, 2004, Ruggerone et al., 2019). However, these results stem from the extremely high variability in abundance of pink salmon in the Gulf of Alaska. For coastal Oregon and Washington, competition among salmon species is believed to occur, but the data to document the magnitude of its impact are not available. Therefore, we do not know the extent to which this ecological process influences early ocean survival, and this remains an important data gap.

We have clear evidence that consumption during early ocean migration is often sufficient for high positive growth rates (Morgan et al., 2018). This might suggest that salmon are not limited by food resources, as might be the case under strong competition. On the other hand, growth varies significantly among years (Dale et al., 2017, Daly et al., 2017), suggesting that, at least in some years, food is limiting. We also know that during poor ocean conditions when prey resources are lower, salmon do not grow as well as fish found in good ocean conditions (Daly & Brodeur, 2015). Therefore, under conditions where competitor abundance is high (either inter- or intra-specific), competition may influence growth rates and may, in turn, influence size-dependent predation rates.

Due to the complexity of this topic and the current data gaps, we are not proposing a direct study of competition at this time. However, much of the data we propose to collect will provide a vital base for potential future work on competition. Moreover, with the ecosystem approach and modeling system we are proposing, we may be able to identify likely competitors and put bounds on the magnitude of competition’s impact.

Predator Species

As described above, salmon smolts typically have access to high quality prey and grow quickly in the marine environment. Of note, in over 20 years of sampling, we have caught only a handful of emaciated salmon, and as such we do not have evidence that salmon experience direct starvation mortality. We therefore hypothesize that predation is one of the strongest drivers of interannual differences in marine survival (Emmett et al., 2006, Wells et al., 2017), though bottom-up drivers can influence the size-dependence of predation mortality (Tucker et al., 2013, Woodson et al., 2013). Unfortunately, predation is the process we know the least about.

Based on 6 years of sampling, fish predators, such as hake, primarily consume forage fish such as anchovy, sardine, herring, and smelt (Emmett et al., 2006); juvenile salmon tend to be consumed incidentally. However, the high abundance of some top piscivorous fishes can result in large numbers of salmon eaten (Emmett et al., 2006). These predators exhibit highly dynamic movement patterns, both seasonally and annually, resulting in high variability of potential impact on salmon populations (Brodeur et al., 2014).

Similarly, avian predators are highly mobile, extremely abundant, and vary at multiple spatial scales (Zamon et al., 2014, Phillips et al., 2017). Data collected by JSOES over the past 16 years demonstrates that two marine bird species, common murres and sooty shearwaters, account for over 70% of all bird observations. Furthermore, these species have distributions that are concentrated around the mouth of the Columbia River during May and June (Zamon et al., In prep.). The few data sets that exist describing marine bird diets suggests that these predators consume enough juvenile salmon that they can be a strong driver of early ocean survival (Varoujean & Matthews, 1983, Wells et al., 2017, Warzybok et al., 2018). Although bird distributions have recently been described in relation to the Columbia River plume (Phillips et al., 2017, 2018), we unfortunately have no information on how avian diets change with environmental conditions. This marks one of the primary data gaps we have and is an important focus for parameterizing an ecosystem model.

Marine mammals primarily consume adult salmon, but juveniles are also eaten, particularly by harbor seals (Chasco et al., 2017a). Estimates of salmon consumption by marine mammals are mostly recent and relatively imprecise. As with bird predators, refining our understanding of abundance, distribution, and diet of these animals would greatly improve our ability to incorporate their effects in management scenario planning for salmon.

An important aspect of understanding predator impacts is the spatial and temporal overlap between predators and salmon. At annual scales, factors such as fish population dynamics and bird migration patterns can drive predator abundance in the nearshore environment. However, there are dramatic intra-annual variations in both distribution and abundance of juvenile salmon and their predators that may confound interpretations of inter-annual fluctuations. Seasonally changing environmental conditions, such as ocean temperatures, salinity gradients, and upwelling can alter predator movements. In addition, many salmon stocks out-migrate at specific times of the year and exhibit specific migration patterns, some of which depend on environmental conditions (Burke et al., 2014), resulting in a highly dynamic spatial and temporal overlap with predators. Importantly, this high inter- and intra-annual variability could be the cause of poor correlations between predator abundance and salmon survival. For example, if hake abundance is particularly high in a given year, but their spatial distribution does not overlap salmon or their age structure is such that most of them are feeding on other resources, the impact of hake may be quite small. On the other hand, even a small number of common murres (for example), if directly overlapping salmon in space and time, can have a large impact on cohort level survival, particularly if alternate prey such as anchovy are scarce (Phillips et al., In prep., Zamon et al., In prep.).

Salmon

Salmon are unique in the coastal environment for several reasons. They are located in the middle of the food web, preying on zooplankton and smaller fishes, but also providing important biomass to the upper trophic levels. They are spatially tied to their rivers of origin, providing an interannually consistent food source for birds, mammals, and piscivorous fish. They also transport biomass from the freshwater environment to the marine environment as smolts, and back to freshwater as adults, linking habitats like few animals can.

The ecological processes described in previous sections vary seasonally, annually, and at longer time scales. Each year, unique conditions drive these processes, resulting in slightly different impacts on salmon growth and survival. Importantly, the drivers of salmon mortality also vary over time and space, such that cohort strength can be set in the estuary, in the Columbia River plume, the nearshore environment, or even during the first winter at sea over a much broader spatial extent (Beamish & Mahnken, 2001). Sampling a salmon population at multiple times and places can aid in estimating when and where mortality events occur. Data derived directly from sampling salmon also provide early warning indicators of ocean conditions and allow early planning for adult salmon return abundance (Burke et al., 2013b, Peterson et al., 2014), and will therefore remain a critical part of our research and monitoring.

Presently, managers base forecasts of salmon returns on the returns of “jack salmon” to streams after their first summer at sea. The number of precocious males that return is more-or-less proportional to the returns of all adults one year later. The predictor has a reasonably good amount of skill because the return of jacks that went to sea when ocean conditions were good tend to have higher survival (i.e., high survival of jacks and older salmon) and lower when ocean conditions were poor. Although the use of counts of jack salmon often work well as an estimator of adult returns, it also fails when mortality events shift in time or when maturation rates change. This can occur at decadal scales as well as interannually, making reliance on this method risky.

Another problem with the jack index is that such data do not provide any understanding of the underlying processes that link salmon survival with ocean conditions. Logerwell et al. (2003), Scheuerell and Williams (2005), and Greene et al. (2005) provide physical descriptive models that hindcast and forecast future returns for coho salmon (Oregon Production Index, OPI, which includes Columbia River coho salmon), Snake River Spring Chinook salmon, and Skagit River fall Chinook salmon, respectively. In each case, physical features such as sea surface temperature, strength of upwelling, and the date of the spring transition were shown to be important attributes that correlated with returns of salmon. These efforts have proven to be valuable first steps, yet each model has since failed to provide accurate forecasts for longer than a few years after being published. Similarly, recent forecasts using sibling regression have over-predicted returns due to a change in maturation rates, possibly due to altered ocean conditions (Brian Burke, NOAA Fisheries, unpublished data). Refined estimates of the ecological processes influencing salmon growth and maturation should be incorporated into sibling regression models, improving their usefulness for tactical management decisions.

A quantitative tool that captures the physical, biological, and ecological processes at play can be used for short-term decisions, but long-term forecasts must also include potential alterations due to climate change. Forecast model failures may be due to mis-specified relationships, non-linear or threshold responses of biological systems to apparently simple linear physical interactions (Hsieh et al., 2005, Samhouri et al., 2017), and over longer time scales, changes in the relationships themselves (Litzow et al., 2018). 

Prioritization

In a budget-limited situation, having a system for prioritizing research and monitoring efforts is particularly useful. Using the framework outlined by Wells et al. (In prep.), we used a series of 14 research themes that are pertinent to salmon ocean ecology to prioritize our efforts (Figure 1; Table 1). But how do we choose which of these are most important? We used two criteria to highlight research themes that should be included in our integrated ecosystem study of marine juvenile salmon. First, there should be some link between the research theme and a management need, though this link does not have to be direct. For example, understanding common murre diet does not link directly to a salmon management decision, but understanding the source of salmon mortality and how this varies from year to year can aid in decisions such as hatchery release timing or serve as an early warning of low (or high) adult returns, impacting harvest management.

Second, we prioritized research themes that are close (in terms of number of links in Figure 1) to salmon (that is, they have a more direct mechanistic link). Currently, many harvest managers consider the PDO as a measure of ocean conditions. Based on Figure 1, there are at least five research themes (arrows) required to connect the PDO to salmon. Each of these comes with uncertainty and interannual variability. Simply admitting that the relationship between PDO and salmon survival is correlative does not address the fact that changes in any ecosystem process between these two components could alter the correlation. As we move towards ecosystem-based science and using mechanistic understanding to manage fisheries, a primary benefit is that fewer assumptions about static relationships are required. As demonstrated by the recent warm blob, a positive (warm) PDO value can represent anything from good conditions for salmon to very bad conditions for salmon. In contrast, a measure of high predator abundance has a more direct link to salmon and should therefore be more tightly linked to salmon survival (making it a higher priority for research and monitoring). Unfortunately, the ecosystem components more closely linked to salmon (prey, competitors, predators) are also the most difficult and expensive to sample. Because of this, they are also the ones we currently know the least about.

We combined these two criteria (a link to management and a mechanistic relationship with salmon) to narrow the list of important research themes in Table 1 to aid in creating a set of research and monitoring efforts that are technically and financially feasible. We describe these sets below in the form of three project options.

Options for future work

We acknowledge the uncertainty in funding facing the region, and BPA specifically. Unfortunately, we are presently underfunded for our current scope of work and have necessarily decreased our personnel by almost half since 2012. We must address this discrepancy between scope of work and funding resources prior to fiscal year 2020. Given this, we put forward three scenarios describing the potential scope of work and rough estimates of the associated budgets. Our goal in this effort is to characterize the landscape of possibilities and get feedback from the region (NWPCC, NWFSC, BPA, etc.) regarding appropriate priorities and scope. We put the three options in increasing order of funding requirements: A) a flat budget, consistent with the reduced budget of the last several years (~$1M/yr), but associated with a reduced scope of work, B) maintenance of the current level of effort/scope, while restoring some of the previously lost funding (~$1.5M/yr), and C) fully restore previous funding and enable the scope of work necessary to effectively monitor the ocean ecosystem and fill our current data gaps for an ecosystem approach to salmon management (~$2M/yr). In all cases, budget requests are in addition to support provided by NOAA (cost share) and other outside agencies.

Table 2. Summary of the three funding options, with a description of the work to be performed and the Objectives addressed. Also included are the additional staffing requirements under each option. Items in blue in rows 2 and 3 represent activities performed in 2018, items in black are new. Activities that will no longer be pursued are not listed.