Heatwave‐induced synchrony within forage fish portfolio disrupts energy flow to top pelagic predators

Abstract During the Pacific marine heatwave of 2014–2016, abundance and quality of several key forage fish species in the Gulf of Alaska were simultaneously reduced throughout the system. Capelin (Mallotus catervarius), sand lance (Ammodytes personatus), and herring (Clupea pallasii) populations were at historically low levels, and within this community abrupt declines in portfolio effects identify trophic instability at the onset of the heatwave. Although compensatory changes in age structure, size, growth or energy content of forage fish were observed to varying degrees among all these forage fish, none were able to fully mitigate adverse impacts of the heatwave, which likely included both top‐down and bottom‐up forcing. Notably, changes to the demographic structure of forage fish suggested size‐selective removals typical of top‐down regulation. At the same time, changes in zooplankton communities may have driven bottom‐up regulation as copepod community structure shifted toward smaller, warm water species, and euphausiid biomass was reduced owing to the loss of cold‐water species. Mediated by these impacts on the forage fish community, an unprecedented disruption of the normal pelagic food web was signaled by higher trophic level disruptions during 2015–2016, when seabirds, marine mammals, and groundfish experienced shifts in distribution, mass mortalities, and reproductive failures. Unlike decadal‐scale variability underlying ecosystem regime shifts, the heatwave appeared to temporarily overwhelm the ability of the forage fish community to buffer against changes imposed by warm water anomalies, thereby eliminating any ecological advantages that may have accrued from having a suite of coexisting forage species with differing life‐history compensations.


| INTRODUC TI ON
The Pacific marine heatwave (MHW) of 2014-2016 was among the largest and most persistent MHWs documented globally to date.
It was characterized by ocean temperatures >2 SD above normal over an area spanning two large marine ecosystems and sustained across multiple winters (Bond et al., 2015;Di Lorenzo & Mantua, 2016;Hobday et al., 2018). The MHW had formed in the upper 100 m of the water column over the central Gulf of Alaska (GOA) by January 2014 (Freeland & Whitney, 2014), had reached coastal regions by the summer, and was largely confined to the inner shelf by September that year (Walsh et al., 2018). The warm water anomaly had deepened across the water column by early 2015 and was enhanced during the winter of 2015-2016 by one of the strongest El Niño events on record (Joh & Di Lorenzo, 2017).
Large-scale mortality events by marine predators and pelagic food web specialists were among the first visible indicators of a major ecosystem perturbation in the GOA during the MHW Savage, 2017) and suggested a major disruption in energy transfer through the middle trophic level.
Forage fish, which play a key intermediary role in energy transfer from plankton to predators, are small pelagic species generally known for "boom and bust" population cycles in which repeated collapse and recovery can occur over long time periods (McClatchie et al., 2017). Similar to a collection of assets in a financial portfolio (Figge, 2004;Schindler et al., 2010), diversity within forage fish communities reduces the variance of the ecosystem services they provide to pelagic marine food webs. Asynchrony within the forage fish portfolio tends to dampen the effects of large fluctuations in abundance by any one species at a time. Niche differentiation, which can arise from decreased interspecific competition or decreased correlations among species-specific responses to environmental conditions, tends to increase asynchrony in population trajectories among species and leads to greater stability within communities (Loreau & De Mazancourt, 2008). Portfolio effects, which provide a measure of the covariance among species within a community, can be used as a framework for understanding stability in ecosystem dynamics over time (Schindler et al., 2015;Thorson et al., 2018).
Major disruptions in energy transfer through forage fish communities are rare because diversity of species with different life-history strategies promotes resilience in marine food webs (Kondoh, 2003;Perry et al., 2010). For example, variability among forage species in migration, age at maturity and seasonality of life stages (Figure 1), and timing of energy allocation toward growth and storage (Heintz et al., 2013;Norcross et al., 2001) can help to mitigate forage fish predators against bottlenecks at middle trophic levels.
Life histories of the three main GOA forage species, Pacific capelin (Mallotus catervarius), Pacific herring (Clupea pallasii), and Pacific sand lance (Ammodytes personatus), hereafter capelin, herring, and sand lance, respectively, are fairly well known. Capelin migrate from the continental shelf to spawn along beaches over protracted periods from spring through fall (Arimitsu et al., 2008;Brown, 2002).
Capelin larvae overwinter, a strategy that favors transformation into their juvenile form at a relatively large size, and development during the peak of productivity in spring (Doyle et al., 2019). It also allows for growth and lipid storage in summer/fall periods at age-1 ( Figure 1). Herring is also a migratory species, as juveniles are reared inshore but move offshore prior to returning as adults, primarily at age-3+, to spawn during spring. Their larval development occurs rapidly during spring, and transformation to their juvenile state occurs by their first summer at age-0 (Doyle et al., 2019). Growth of juveniles during summer and fall is critical to successful overwintering of herring (Heintz et al., 2013;Sewall et al., 2019). Sand lance is a non-migratory species associated with nearshore sandy bottom habitats (Baker et al., 2018;Ostrand et al., 2005) that spawns from fall to winter months (Robards, Anthony, et al., 1999;. Hatch timing in sand lance is mediated by temperature, with larvae undergoing rapid growth during spring (Doyle et al., 2019). Like herring, the energy allocation strategy favors juvenile growth of sand lance during the summer and fall prior to their first winter at age-0, and subsequently switches to lipid storage during age-1 (Robards, Anthony, et al., 1999;. Although any forage fish life stage can be targeted, adult spawning aggregations tend to concentrate marine predators in space and time (Womble et al., 2005).

F I G U R E 1
Generalized life-history strategies, including monthly timing of spawning, larvae, and early juvenile stages, for three key forage fish species in the North Pacific In this study, we investigate changes within the forage fish community that manifested in the northern GOA during the MHW, including variation in abundance and changes in age structure, size, growth, and energy content of species common in diets of pelagic predators. Incorporating new data from long-term monitoring efforts within the region, we put these forage fish observations into context with coincident changes in zooplankton and upper trophic level predators. Finally, with a focus on the pelagic trophic pathway in the GOA, we identify potential mechanisms that underlie mass mortalities and reproductive failures in top predators that were associated with the MHW (Barbeaux et al., 2020;Piatt et al., 2020;Savage, 2017).

| Study area
The study area includes the northern GOA, Prince William Sound (PWS), Cook Inlet, and the offshore region between 143-156°W and 56-61°N ( Figure 2a). Circulation in the coastal and shelf regions of the northern GOA is influenced by the Alaska Coastal Current, the dominant alongshore current, as well as wind mixing and freshwater input (Stabeno et al., 2016;Weingartner et al., 2005). In PWS, with basin depths exceeding 700 m, the Alaska Coastal Current also defines transport from east to west through two ocean entrances in PWS, although similar to conditions on the shelf, circulation is mediated by seasonal and interannual variation in winds and freshwater runoff (Niebauer et al., 1994). Dynamic ocean conditions characteristic of the region are driven in part by complex bottom topography on the shelf and coastal embayments (Figure 2a). Strong winter downwelling relaxes seasonally, which enriches nutrient supply to the shelf and promotes spring and fall bloom activity that are important in driving productivity in this system (Childers et al., 2005; Waite & Mueter, 2013).

| MHW indices
Marine heatwave events within in the study area were defined as anomalously warm events lasting five or more days, with sea surface temperatures (SST) warmer than the 90th percentile (Hobday et al., F I G U R E 2 (a) Map of the study area, including sampling locations in Prince William Sound (PWS) and the northern Gulf of Alaska (GOA). Capelin survey extent includes the larger Gulf of Alaska region (blue shaded area, inset). Seafloor depth reaches over 4500 m in the study area (>700 m in PWS) with lighter shading representing shallower depths, the 500 m isobath is denoted by a dashed-blue line. Distribution of marine heatwave event indices during 2014-2016 including (b) cumulative number of heatwave days, and (c) mean of long-term  sea surface temperature anomalies for heatwave events in each pixel 2016). Daily ¼ degree SST data (NOAA OISST 1982 Table 1; Huang et al., 2020) from the northern GOA region were extracted and analyzed using the R package "heatwaveR" (Schlegel & Smit, 2018). To identify the spatial extent, duration, and magnitude of the MHW, we summarized the total number of heatwave days and the mean intensity (temperature anomaly, °C) of events within each grid cell.

| Seabird diets
Information on predator diets are among provide some of the longest time-series data for most forage species in the region (Hatch & Sanger, 1992;Piatt et al., 2018;Sydeman et al., 2017). Seabird diet samples were collected from adults provisioning chicks during the breeding season (June-August) from 1993 to 2018 using methods described previously (Hatch, 2013(Hatch, , 2019Schoen et al., 2018;Thayer et al., 2008). Frequency of occurrence in diets of obligate surface feeders (black-legged kittiwakes, Rissa tridactyla) and divers (rhinoceros auklets, Cerorhinca monocerata, and tufted puffins, Fratercula cirrhata) provide indices of availability for preferred prey species  within the approximate foraging range that encompasses much of the continental shelf and coastal waters around the colony at Middleton Island (Table 1, Figure 2a; Cunningham et al., 2018;Osborne et al., 2020). For each seabird feeding guild, availability indices (frequency) were calculated by forage fish species and year.
We used portfolio effects as a measure of synchrony among frequencies of these preferred forage fish species for each seabird feeding guild (Barnes et al., 2020;Loreau & de Mazancourt, 2008;Thorson et al., 2018). To identify patterns across relatively short time-scales, we calculated variance ratios VR i , j for each of i 3-year centered rolling time windows and j feeding guilds such that: is the variance of the frequencies summed across forage fish species at time window i, for feeding guild j, and ∑ var is the sum of the variance of frequencies summed across forage fish species at time window i for feeding guild j. Portfolio effects were calculated as 1 − VR i,j so that lower values indicate greater synchrony across species and therefore lower trophic stability in the system (Thorson et al., 2018).

| Survey indices
Most forage fish species in the region have not been commercially exploited in the region owing to conservation measures put in place in the late-1990s (Witherell et al., 2000), and therefore directed surveys to inform management of these species are limited. The exception is herring in PWS, which had historically been fished for more than a century but has been closed since 1999 due to insufficient biomass (Muradian et al., 2017  Age was assigned by counting translucent zones in sagittal otoliths. To reduce variability associated with sexual dimorphism in spawning capelin, we identified changes in the distribution of length at age (total length, mm) for male capelin in spawning condition based on external morphology (i.e., raised lateral line and enlarged anal fin).
was assessed with bomb calorimetry. Whole-body total energy (kJ) was estimated by multiplying the energy density (kJ g −1 dry mass) by dry mass (g), and then compared among years with ANOVA and Tukey's HSD post hoc test (α < 0.05).
Because age data were not available for the entire time series, we used length data (total length, mm) from sand lance sampled by seabirds at Middleton Island to identify changes in size structure over time. Diving auklets and puffins deliver whole and undigested prey samples to their chicks, which allows for measurements of fish length and mass (Thompson et al., 2019). We used lognormal mixture models (MacDonald & Pitcher, 1979) to estimate the proportion of multimodal size classes with annual length frequency data and multiplied proportions for each size class by the CPUE in each year. The first size group contained fish <100 mm in length and typically corresponds to age-0 fish in summer. Age classes of the second and third size groups could not be determined reliably from length (von Biela et al., 2019), but are assumed to be between age-1 and age-4 (Robards, Anthony, et al., 1999;. Analyses were conducted with the R package "mixtools" (Benaglia et al., 2009).

| Herring growth and energy content
Scale growth increment measurements were used to document long-term patterns of growth in herring from 1978 to 2017 (Table 1; Alaska Department of Fish & Game, 2019; Batten et al., 2016).
Age, sex, and length data were collected from commercial fisheries (1978)(1979)(1980)(1981) and fishery independent (1982-2017) research projects outlined by Baker et al. (1991). Samples were collected with purse seines and cast nets periodically through the spawning season, targeting a sample size of 450 fish for measurements of standard length (mm), weight (g), and sex. Maturity was determined from the gonads. Scales were mounted on slides, annuli counted to determine age, and archived by the Alaska Department of Fish and Game. A subset of these scales was imaged, with a target of 180 scales (30 each from males and females age-4 through age-6) from each year. Scales were photographed or scanned at known magnification, and the width of growth increments were measured digitally with Image-Pro software, to determine the distance between adjacent annuli. Age and sex were not significantly related to growth (ANOVA, p = 0.82 and 0.69, respectively) so we pooled the data for further analysis. and processed for energetics using bomb calorimetry and carbon to nitrogen ratios (Gorman et al., 2016(Gorman et al., , 2018 year. We then estimated least square means of back-transformed total energy for each year. For these analyses, we used "lme4" and "emmeans" packages in R version 4.0.2 (Bates et al., 2015;Lenth, 2020; R Core Team, 2020).

| Zooplankton abundance indices
To identify coincident changes in seasonal zooplankton standing stock during the MHW, we summarized trends in euphausiids and copepod abundance indices from several sources (Table 1) (September). These samples were collected at night using a 500-µm mesh Hydrobios multinet system towed from 100 m depth (Sousa et al., 2016). To identify spatial differences along the GOA shelf and in PWS, we summarized the station-level data by season and region: PWS (stations within the large embayment), and Inner Shelf region (shore to 50 km from shore), which is heavily influenced by the Alaska Coastal Current; the transitional Middle Shelf region (50 km from shore to the shelf-break, defined using the 1000 m depth contour); and the Oceanic region (slope and basin, beyond the 1000 m depth contour).

| Copepods
We summarized patterns in the seasonal abundance of copepods using monthly data from CPR (Batten et al., 2018)

| Acoustic macrozooplankton
We also classified macrozooplankton along acoustic transects in PWS during September using multifrequency acoustic methods.
Acoustic surveys were conducted using Simrad EK60 (Kongsberg Maritime) echosounder with a towed (2014,(2017)(2018) or hull mounted (2019) split-beam transducers (38 and 120 kHz, 12° and 7° beam widths, respectively, measured at half-power) along transects in three sub-regions that typically contain foraging aggregations of humpback whales (Megaptera novaeangliae; Figure 2a). The echosounder system was calibrated during each survey by suspending a 38.1-mm tungsten carbide sphere following standard practices (Foote et al., 1987). Acoustic backscatter in the water column were classified for depths between 4 m below the surface and 1 m above the bottom using frequency response methods for

| Marine predators
We identified changes in distribution and abundance of two conspicuous marine predators that experienced mass mortalities as one of the first conspicuous signs that an ecosystem disruption was underway during the MHW Savage, 2017).

| MHW duration, magnitude, and extent
During 2014-2016, MHW events summarized within each ¼ degree resolution grid cell in the northern GOA ranged in cumulative duration from 291 to 813 days (mean ± SD: 646 ± 67 days).
The mean intensity across all events within a given grid cell

| Forage fish availability in seabird diets
Across all years, 79% (n = 5281) of diver diet samples and 76%   (Figure 3a). Herring frequency anomalies were not consistent between surface feeder and diver diets for more than a decade prior to the MHW but surface feeders in particular relied on herring to a greater extent during and after the MHW, which is consistent with inshore feeding and greater foraging distances by kittiwakes during the MHW (Osborne et al., 2020).
When these time series of forage fish indices from seabird diets were summarized by the portfolio effects (1-variance ratios), a rapid decline at the onset of the MHW identified strong trophic instability and synchronous collapse of the forage community in the northern GOA in 2014-2015 (Figure 3b). More stable, asynchronous dynamics across forage fish species persisted for at least two decades prior to the MHW, and returned shortly after the MHW dissipated. Similar patterns in portfolio effects between surface feeding and diving seabirds suggest a common underlying mechanism, although synchrony among species in diver diets was more pronounced than it was for surface feeders.

| Forage fish trends in survey indices
Survey

| Zooplankton indices
In general, euphausiid biomass along the GOA shelf and in PWS was dominated by Thyanoessa inermis and T. longipes from 2012 to 2014 and mean biomass of these species was reduced abruptly by 2015 ( Figure 7). Changes in mean biomass of euphausiids, therefore, were driven by the loss of these species, particularly during spring.
Within the bays of PWS abundance anomalies of the dominant large copepod Neocalanus spp. were as much as 2 SD below the mean in spring 2015, and >1 SD above the mean in spring 2016.
Resident copepods and warm water copepods had inverse trends, which signaled a sustained community shift due to the influx of warm water copepods that occurred especially in summer and fall

| Marine predators
Marine predators specializing on pelagic forage species in the northern GOA experienced shifts in distribution, reduced encounter rates, and reduced reproduction. PWS humpback whale encounter rates increased on fall surveys through 2014 and declined F I G U R E 6 Prince William Sound (PWS) herring indices, (a) age-3 growth anomalies from scale increment width measurements, (b) spring pre-spawning weight anomalies, and (c) estimated least square mean (±95% CI) juvenile total energy and mean energy density (ED, kJ g −1 wet weight, color) in March and November. Anomalies were standardized to mean 0, SD 1.

| DISCUSS ION
A primary mechanism of trophic disruption in the multi-year MHW was that it reduced the abundance of several key forage fishes and some cold-water-associated euphausiids (T. inermis and T. longipes) while it simultaneously lowered nutritional quality of many forage species. A steep, but short-lived decline in portfolio effects at the onset of the MHW points to strong trophic instability in the system due to synchronous collapse of the forage fish community. Diversity of life-history strategies, including differential tolerances to water temperature and seasonality of spawn timing or growth phases, typically promotes staggered population trajectories of forage species; however, this mechanism for creating stability in marine food webs apparently failed during the MHW.
The MHW was in fact a relatively brief heating event compared to the 1977 North Pacific Regime Shift, which was driven by the Pacific Decadal Oscillation and involved a similar temperature range F I G U R E 8 Monthly log-transformed copepod abundance indices (color) standardized anomalies (bars) for species groups sampled in Prince William Sound (PWS, bongo net, m −3 ) and along the Gulf of Alaska shelf (continuous plankton recorder, sample −1 ). Dotted horizontal lines demarcate the timing of the Pacific marine heatwave, and spring months (March-May) are shaded in light blue in its transition from cool to warm anomalies but played out over more than a decade rather than just 2 years (Anderson & Piatt, 1999;Hare & Mantua, 2000;Litzow & Mueter, 2014). This climate oscillation led to a complete reorganization of pelagic food webs in the GOA and restructured marine communities, from phytoplankton to whales (Chavez et al., 2003;Francis et al., 1998). Despite the magnitude of biological change wrought by regime shifts, which by definition imply large-scale reorganizations of marine ecosystems (Overland et al., 2008), those changes are not generally evidenced in abrupt and short-term impacts such as wholesale recruitment failures, precipitous population declines, or mass die-off events (Chavez et al., 2003;Litzow & Mueter, 2014;Yatsu et al., 2008).
So, despite its relatively brief tenure in the NE Pacific, the MHW elicited an abrupt response among many members of the pelagic food web. In this respect, "heatwaves" appear qualitatively different from "regime shifts" (Litzow et al., 2020), and this begs for a mechanistic explanation of how and why heatwaves disrupt food webs. Species-specific thermal optima are not well known, but because capelin is a more northern species associated with cooler ocean temperatures (Sydeman et al., 2017;Thompson et al., 2019) it is likely that under warming conditions their distribution contracted or shifted locally to deeper, cooler waters (Arimitsu et al., 2008;McGowan et al., 2019;Rose, 2005). has been demonstrated in other species (Laurel et al., 2016).

| Truncation of age and size structure in forage fish
Truncation of the older, larger size classes in capelin and sand lance, both species with relatively short life spans, could have been due to top-down pressures during the MHW. Changes in the demographic structure of fishes are a well-documented consequence of size-selective removal processes such as exploitation due to predation, disease, or fishing (Ohlberger et al., 2014;Perry et al., 2010). Such size truncation reduces the buffering capacity of the population to environmental variability leading to poor recruitment and survival Planque et al., 2010). It can also alter spawning dynamics. For example, smaller, younger individuals produce fewer eggs, since fecundity is strongly related to size (Gjøsaeter, 1998), and spawning by younger populations may occur over a reduced duration and spatial extent .
In the absence of forage fish fisheries in the region, top-down pressures on forage species during the MHW are hypothesized to F I G U R E 1 0 Murre (Uria spp.) standardized density anomalies (mean of 0 and SD of 1 for each season and region). September surveys in the oceanic region, in which no murres were observed, are indicated by 'x'. Dotted horizontal lines demarcate the timing of the Pacific marine heatwave. PWS, Prince William Sound have primarily resulted from increased metabolic needs of large predatory fish over multiple years of warm conditions (Barbeaux et al., 2020;Piatt et al., 2020). A 2°C increase in temperature would have increased the daily rations of GOA pollock, Pacific cod, and arrowtooth flounder (Astheresthes stomias) by 70%, 34%, and 65%, respectively (Holsman & Aydin, 2015). Moreover, owing to a strong year class in 2012, the walleye pollock biomass in the region nearly doubled in the GOA between 2013 and 2015 (Dorn et al., 2019), which may also have led to a rapid increase in predation pressure on prey species. The sustained warm temperatures spanning two winters, along with increased consumption needs by ectothermic predatory fish, could have provided a mechanism for large-scale top-down control of the middle trophic level.
Disease has also been identified as a top-down stressor limiting the PWS herring population. The primary pathogens are Viral Hemorrhagic Septicemia virus, with infection rates increasing when schools occur in dense aggregations such as during periods of spawning or predator-corralling, and Ichthyophonus, which tends to be more prevalent with increasing host size and age (Hershberger et al., 2010(Hershberger et al., , 2016. No relationship between warming temperatures and disease prevalence has been established, but disease is considered an underlying stressor that may be a factor in the failure of the PWS herring population recovery.  (Tomiyama & Yanagibashi, 2004), whereas juvenile herring remain in the water column to feed, particularly in warm temperatures when metabolic rates are higher (Sewall et al., 2019).

| Decreased quality of forage fish
TA B L E 1 Description of datasets used to identify changes in availability, abundance, and quality of forage species, as well as coincident changes in lower and upper trophic-level taxa in the Northern Gulf of Alaska (GOA) during the 2014-2016 Pacific marine heatwave.

| Shifts in zooplankton communities
Euphausiid biomass and macrozooplankton abundance indices decreased in the region after 2014 (this study). Lower euphausiid biomass was driven by the loss of T. inermis and T. longipes, both species that are associated with cooler temperatures (Iguchi & Ikeda, 2005;Pinchuk et al., 2008). Pinchuk et al. (2008) found T.
inermis abundance was positively associated with a cool regime in 1999-2002, perhaps because the species has a relatively high mass-specific metabolic rate thus spends significantly more energy on metabolism during warmer winters. Following a warm winter, T. inermis would have less energy reserves in spring for spawning.
While euphausiids declined in biomass during the MHW, copepod biomass generally increased on the shelf, which is likely related to faster copepod development rates in warmer water (Kimmel & Duffy-Anderson, 2020

| Impacts to pelagic predators
The magnitude of this system-wide perturbation was evidenced in a coherent response of several dominant top predators within each of the important marine vertebrate classes. In birds, for example, there was a mass-starvation die-off of an estimated 0.5-1.2 million common murres in the NE Pacific during 2015-2016, widespread total reproductive colony failures in 2015-2018, and mass mortalities of several other seabird species Piatt et al., 2020).
Although the murre die-off occurred along the entire US west coast, densities were greatest in PWS and coastal regions along the GOA during winter 2015-2016  following a marked redistribution of the species to inside and coastal waters (this study).
In mammals, there was mass mortality of humpback and fin whales in the GOA during 2015-2016 (Savage, 2017), as well as precipitous declines in abundance and calf production rates of humpback whales throughout the region (this study, Neilson & Gabriele, 2019).
Similarly, in fish, Barbeaux et al. (2017) showed there was a large (>71%) reduction in biomass of Pacific cod from 2015 to 2017 resulting in a severe (77%) reduction in cod fishing quotas. Moreover, increased mortality and diminished body condition of Pacific cod during the MHW were attributed to increased metabolic demands and nutritional stress during the protracted warm conditions (Barbeaux et al., 2020).

| Implications
The large spatial scale in which mass starvation of large-bodied pelagic specialist predators was associated with the MHW suggest our aggregate findings, despite the disparate spatial and temporal domains within the datasets compiled here, are representative of the larger GOA marine ecosystem. This information provides important context in understanding the response of large marine ecosystems to current and future events. For example, similar to our findings, recent severe heatwave conditions across both hemispheres have been associated with zooplankton community shifts, species range expansions, and unusual mortality in pelagic predators (Evans et al., 2020;Romano et al., 2020;Ruthrof et al., 2018).
Marine heatwaves such as this are associated with climate change and are increasing in frequency and duration globally (Laufkötter et al., 2020;Oliver et al., 2018). The 2014-2016 MHW caused a large-scale disruption of the pelagic food web, manifested through a simultaneous reduction in the abundance and quality of key forage taxa. The MHW appeared to overwhelm a generally diverse and resilient forage community, leading to catastrophic failures in top pelagic predators and eliminating any ecological advantages that might have accrued from having a suite of coexisting forage species with differing life-history compensations ("portfolio effect").
Changes at lower and upper trophic levels provide evidence for a large-scale reduction in energy transfer through the middle trophic level (forage fish and euphausiids) during the MHW. Declines in the abundance and quality of forage fish and euphausiids were likely a primary driver of the mass mortalities, changes in distribution, reproductive failure, and malnutrition observed in piscivorous marine predators that depend on the pelagic food chain.
The ecosystem response observed appears fundamentally different from temperature-induced "regime shifts," which transition between cold and warm phases without accompanying disruptions in energy flow, as evidenced by large-scale mortality of seabirds, whales, and predatory fish. The abrupt but brief decline in portfolio effects by the forage fish community during this event appears to differentiate ecosystem responses to "regime shift" versus "heatwave" temperature anomalies and warrants further investigation for understanding and anticipating the response of large marine ecosystems to future heatwave events.

ACK N OWLED G EM ENTS
We thank the scientific and program management team of the Gulf Watch Alaska long-term monitoring (https://gulfw atcha laska.org/), Herring Research and Monitoring (https://pwssc.org/herri ng/) programs, and state (Alaska Department of Fish and Game) and federal (NOAA Alaska Fisheries Science Center) fisheries management agencies for their expertise and resources in support of this work.
We thank the many field crews that contributed to decades of data collection efforts. This research was supported by the Exxon

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly avail-