Multispecies population-scale emergence of climate c hang e signals in an ocean warming hotspot

Ocean waters of the Northeast US continental shelf have warmed rapidly in recent years, with sea surface temperatures rising 2.5 times faster than those of the global oceans. With this strong warming trend, the frequency and duration of marine heatwaves have increased. These temperature changes stood out as a distinct warm temperature regime during the 20 1 0s. During this decade, fish population characteristics also differed from the past. Species distribution shifts were detected for many species, demonstrating one way species could adapt to warming conditions. However, for most species, distribution shifts were insufficient to avoid warmer surface or bottom temperatures. As species occupied warmer habitats, growth patterns aligned with expectations for warming temperatures. Consistent with the temperature-size rule, some species exhibited faster growth at early life stages but plateaued at smaller body sizes; other species, however, experienced reduced growth across all ages, indicating thermal stress. Finally, population productivity indexed by the recruit-to-spawner ratio declined significantly during the 20 1 0s for some populations. Changes in these three processes— distribution, growth, and productivity—indicate the emergence of climate change signals across multiple Northeast US fish populations. These effects create new challenges for fishery managers and industry participants operating in the context of non-stationarity and uncertainty.


Introduction
The Northeast Shelf of the USA is characterized by strong seasonal cycles and interannual variability in oceanographic conditions that drive nutrient availability and energy flows through the ecosystem (Parr 1933 , Pershing andStamieszkin 2020 ).This ecoregion has historically been renowned for its productive fisheries, with the challenges of overfishing dominating management concerns in the region (Rosenberg et al. 2006 ).In recent decades, fishing pressure has begun interacting with a new challenge-climate change-which is affecting fish populations and fisheries management in this region and across the globe (Pershing et al. 2015, Britten et al. 2017, Cheung et al. 2022 ).
Ocean waters of the Northeast Shelf have warmed more rapidly than most of the world's oceans (Pershing et al. 2015 ), placing this region at the forefront of experiencing climaterelated changes and fishery management challenges.Strong warming trends in surface and bottom waters (Pershing et al. 2015, Record et al. 2019, Friedland et al. 2020 ) are increasingly being punctuated by marine heatwaves (Mills et al. 2013, Scannell et al. 2016, Pershing et al. 2018 ).Moreover, changes have developed in seasonal cycles of warming and cooling (Friedland et al. 2015, Thomas et al. 2017 ).These ocean temperature patterns reflect the overall warming of the world's oceans (Alexander et al. 2018 ), but they have been exacerbated in this region by changes and variability in ocean and atmospheric circulation associated with global climate change (Gonçalves Neto et al. 2021, Karmalker and Horton 2021, Meyer-Gutbrod et al. 2021 ).In addition, the region is being affected by other climate-related changes in environmental conditions, including ocean acidification and strengthening stratification (Brickman et al. 2021, Siedlecki et al. 2021 ).
The effects of these environmental changes are permeating through the region's marine ecosystem.The large copepod, Calanus finmarchicus (hereafter, Calanus ), plays a key role in distributing energy across seasons and through the ecosystem (Pershing and Stamieszkin 2020 ).Warming and its influence on stratification have been associated with seasonal changes in Calanus , leading to reduced abundance during the summer and fall and increased abundance in winter (Runge et al. 2015, Record et al. 2019 ).When Calanus abundance is low, a more diverse but smaller-bodied plankton community emerges (Pershing et al. 2005 , Pershing andKemberling 2023 ), which alters energy pathways and affects productivity at higher trophic levels (e.g.right whales: Meyer-Gutbrod et al. 2015 ;cod: Mountain and Kane 2010 ).
Direct and indirect effects of warming on primary and secondary productivity influence higher trophic levels, as observed in fish and invertebrate populations and communities on the Northeast US Shelf.In many marine ecosystems across the world, "fingerprints" of climate change impacts on marine ecosystems are widely apparent, including changes in spatial distribution, seasonal phenology, individual growth, and population productivity (e.g.Poloczanska et al. 2013, Baudron et al. 2014, Britten et al. 2016, Free et al. 2019, Cooley et al. 2022 ).These same types of changes are occurring in the Northeast Shelf marine ecosystem.Prior studies in the region have identified multiple factors that influence species distribution (e.g.Murawski and Finn 1988, Perry and Smith 1994, Wang et al. 2018, McHenry et al. 2019 ), and recent studies have indicated that changes in spatial distribution are occurring across many species as they track climate directionality and velocity (Nye et al. 2009, Pinsky et al. 2013, Kleisner et al. 2016 ).However, some species and functional groups have shifted in directions that are not aligned with preferred habitat conditions (Fuchs et al. 2020 , Thorne andNye 2021 ), and distinct dynamics are emerging at the leading versus trailing edges of species' ranges (Fredston-Hermann et al. 2020 ).In addition to spatial patterns, temporal dynamics are also changing, with shifts in the phenology of life history events documented across many components and levels of the ecosystem (Henderson et al. 2017, Staudinger et al. 2019 ).Further, multiple interacting influences of warming are affecting the recruitment and productivity of fish and invertebrate populations in both beneficial and detrimental manners (Pershing et al. 2015, Le Bris et al. 2018, Tableau et al. 2019 ), although a general decline was detected across multiple species during much of the 2000s (Perretti et al. 2017 ).
In this paper, we provide an updated perspective on ocean warming in the Northeast Shelf region and concurrent changes in fish populations during the 2010s decade.We focus on ocean warming (the long-term trend) and marine heatwaves (shorter-term periods above long-term average temperatures) as substantial manifestations of climate change in the region, and as features that distinguish the 2010s from prior decades.Comparing the 2010s to prior decades, we analyze differences in fish populations based on three processes-spatial distribution, individual growth, and population productivity-that may be influenced by temperature and that are interconnected with one another.We hypothesize that distribution shifts toward the poles or to deeper depths will enable species to buffer the effects of warming by maintaining stable temperatures in the habitats they occupy.The degree to which they effectively maintain preferred temperatures will affect growth.Based on the temperature-size rule, warmer temperatures are expected to result in faster growth at young ages before plateauing at smaller adult sizes (Atkinson 1994 ), and we use this rule as the basis for our hypothesis for growth effects.Subsequently, we hypothesize that growth changes will also affect population productivity.If maximum body sizes decline with warming, reduced fecundity may result in lower per-capita recruitment and declines in productivity (Cheung et al. 2013, Baudron et al. 2014, Perretti et al. 2017 ).Our hypotheses are structured around expected effects associated with warming, and those effects may accrue through direct or indirect mechanisms, including secondary changes in the abundance and quality of prey, altered predator fields or other sources of mortality, and interactive effects with other stressors such as fishing pressure.
Here, we describe patterns in ocean warming and fish population processes in the Northeast U. S. Shelf region and identify associations between multiple physical and biological changes that occurred during the 2010s.Ultimately, we syn-thesize how climate change signals are appearing across multiple fish species in the region and draw insights into ecosystem and fishery implications.We also provide a foundation for future studies to more closely examine patterns, drivers, and mechanisms associated with changes in population processes.This study demonstrates the emergence of climate change signals in both physical conditions and biological processes in the Northeast Shelf marine ecosystem during the 2010s and characterizes elements of a distinct ecosystem regime relative to prior decades.

Sea surface temperature data
Global sea surface temperature (SST) data were obtained from NOAA's Optimum Interpolated SST (OISST v2) product, which provides daily SST values at a 0.25 • latitude x 0.25 • longitude resolution (Huang et al. 2021 ).These data were obtained from the NOAA Physical Sciences Laboratory via https:// psl.noaa.gov/data/ gridded/ data.noaa.oisst.v2.highres.html .A daily climatology for every 0.25 • pixel in the global data set was created using average daily temperatures spanning the period of 1982-2011.Daily anomalies were computed as the difference between observed temperatures and the daily climatological average.Average annual anomalies for 1982 to 2021 were computed by averaging over all pixels in the study region and across the global oceans, and linear regressions were used to determine warming trends ( • C yr −1 ) for each of the annual anomaly time series.Marine heatwaves were computed following Hobday et al. (2016) , which define heatwaves as periods when daily SSTs are above the 90 th percentile for that day in the climatological record .Under this definition, the initiation of a heatwave event requires that temperatures remain above this threshold for at least five consecutive days, and an event ends when temperatures drop below the threshold for more than two days (Hobday et al. 2016 ).

Fish biological data
Biomass, distribution, and growth data were obtained from annual spring and fall bottom trawl surveys conducted by the NOAA Northeast Fisheries Science Center from 1970 to 2019 (Grosslein 1969, Azarovitz 1981, Politis et al. 2014 ).The surveys covered the Northeast US Shelf from Cape Hatteras, North Carolina to the Gulf of Maine, including Georges Bank ( Fig. 1 ).A stratified random sampling design was used, with strata defined based on depth, bottom habitat type, and latitude.Stations were randomly selected within each stratum proportional to the stratum area, with a minimum of two successful stations required per stratum during a survey.At each station, a bottom trawl net was towed along consistent depth contours for a standardized time and speed to sample the fish community; in addition, in situ conditions were measured, including surface temperature, bottom temperature, and depth.We reduced the full data set by using only tows for which there were no gear or duration problems, and we selected offshore survey strata that were consistently sampled over the survey time series ( Fig. 1 ).The allocation of survey sampling effort among strata is more fully described in Supplement S1 .
The catch was sorted to species, counted, and weighed.Abundance and biomass measures at the species-tow level were adjusted using standard calibration factors that account for changes in vessels, gear, and doors over the duration of  the survey (Sissenwine and Bowman 1978, Byrne and Forrester 1991, Miller et al. 2010 ).In addition, individual lengths were measured to the nearest 1 cm for all or a sub-sample based on the amount of catch of each species.Over the range of lengths observed in a tow, biological samples, including individual weights, were measured for a representative subsample of the catch starting in 1992.Biological samples for age determination were stored dry in envelopes, and returned to the lab.Age estimates (years) were determined by observing annuli on various hardparts, typically otoliths, following species-specific procedures established in Penttila and Dery (1988) and conducted as described by NOAA Fisheries (2023a) .

Distribution and growth metrics and analyses
To compute distribution and growth metrics, we curtailed the time series in 2019 due to COVID-related survey disruptions in 2020, and we excluded 2017 from the distribution analyses as mechanical issues prevented sampling of the Mid-Atlantic Bight in fall 2017 ( Supplement S1 ), which may have created northward and eastward biases in distribution metrics.A sensitivity analysis evaluating the effect of oversam-pling one stratum (stratum 16) during 2006-2008 indicated only minor changes to the study results and no changes to major conclusions, and this stratum was included to provide consistency with previous studies of distribution shifts in this region (Nye et al. 2009, Pinsky et al. 2013, Kleisner et al. 2016 ) ( Supplement S1 ).
For distribution analyses, we selected 49 species that met three criteria: (1) present in 90% of surveyed years, (2) present in 80% of years for each season in both periods (1970-2009 and 2010-2019), and (3) present in > 10 tows on average in each season ( Supplementary Table S1 ).For each species, we calculated five biomass-weighted metrics: (1) center of latitude, (2) center of longitude, (3) depth, (4) sea surface temperature, and (5) bottom temperature.These metrics were calculated as a weighted average within season and then averaged across seasons within a year: where X is the parameter of interest (latitude, longitude, depth, surface temperature, and bottom temperature), j is the survey year, and w i is the biomass of the species at each station.Treating each season equally in the annual average represented a species' environmental conditions based on spending half of a year in habitat conditions encountered during each season.While there are considerable seasonal differences in habitats occupied for many species, this averaging approach provided a reasonable approximation of aggregate annual conditions that could be related to growth and productivity patterns.We computed a stratified mean annual abundance time series for each species to qualitatively consider whether spatial distribution patterns were associated with relative population size.
For growth analyses, we used data for 14 species that are routinely sampled and aged by the Northeast Fisheries Science Center ( Supplementary Table S1 ).We restricted our analysis to species for which individual length (cm) and weight (kg) were consistently measured by the biological sampling program.In addition, we curtailed the ages included in the analysis to ensure adequate sample sizes existed in both time periods (average of n > 4 measurements at age per year).Rather than constructing a cohort-based analysis of growth curves since few cohorts would be fully represented at a decadal scale, we focused on age-based analyses of length and weight.
Differences in distribution (latitude, longitude, depth, surface temperature, and bottom temperature) and size-at-age (length and weight) metrics were compared for each species during the decades spanning 1970-2009 and the recent decade of 2010-2019.Recognizing that an assumption of equal variance would be violated for some of the comparisons, we used W elch' s t -tests to compare the two groups of years using the t.test function in R (version 4.2.2).

Population productivity data and analyses
We investigated potential changes in population productivity by examining changes in spawner-recruit relationships based on outputs from regional stock assessments.Through the NOAA Fisheries Stock Status, Management, Assessments, and Resource Trends (SMART) website, we downloaded the most recent stock assessment reports (NOAA Fisheries 2023b ) for stock units of species that were included in the growth analyses, aside from silver hake and witch flounder, for which estimated recruitment values were not available ( Supplementary Table S1 ).From these reports, we compiled a time series of spawning stock biomass (SSB) and recruitment (R) for each stock ( Supplementary Table S2 ).The range of years for which data were available differed across the stocks, varying both in how far back in time the assessment was run and the year of the most recent assessment.We created a productivity index as R/SSB by lagging R as needed to align with the corresponding SSB that produced that year class.
To characterize the 2010-2019 decade, we split the R/SSB time series for each species into pre-2010 and post-2010 (which included 2010) groups and tested for a significant difference using W elch' s t -test as implemented in the t.test function in R (version 4.2.1).We also tested differences in the SSB time series in an identical manner to aid in interpreting and explaining the R/SSB results.Since this analysis is based on stock assessment outputs, the uncertainty associated with the R and SSB estimates is expected to vary over the time series (Brooks and Deroba 2015 ).We do not have access to the error estimates associated with the R and SSB outputs.Instead, we evaluated the potential for R/SSB estimates during individual years to strongly influence the t-test results ( Supplement S2 ).Warming rates have not been spatially consistent over the region, however ( Fig. 2 b).The strongest warming was experienced in the Gulf of Maine, where waters warmed at rates of 0.48 • C per decade since 1982, a rate that is 3.2 times faster than the rate of warming experienced over the world's oceans during that time.In the Gulf of Maine, mean annual SST between 1982-2009 and 2010-2021 differed by 1.379 • C, representing an increase from a mean of 9.82 • C to 11.2 • C in average SST during the two time periods, respectively ( Supplementary Fig. S1a ).Annual warming rates were lower in the Mid-Atlantic Bight and along the southern flank of Georges Bank.

Warming ocean temperatures and marine heatwaves
The warming trend in the region has pushed temperatures further from those experienced during the climatological reference period , resulting in more frequent and protracted marine heatwaves ( Fig. 2 c).A major heatwave occurred in the region in 2012, characterized by strong warm temperature anomalies that persisted over nearly the entire year.Since 2012, marine heatwaves have occurred for some portion of each year, particularly in the summer and fall months, although winter and spring heatwaves occurred in several years.Just as warming rates have been stronger in the Gulf of Maine, so too has been the heatwave record ( Supplementary Fig. S1b ).

Species distribution shifts and temperature exposure
Many of the species we analyzed demonstrated statistically significant differences in latitude, longitude, and/or depth between the 1970-2009 and 2010-2019 periods ( Table 1 ).Of the 49 species included in our analyses, 23 showed northward shifts in latitude that are consistent with a pattern of poleward movement ( Fig. 3 , Supplementary Figs S2 and S3 ; Table 1 ).Some of the largest northward latitudinal changes were observed for black sea bass (1.75 • N), American lobster (1.25 • N), and blackbelly rosefish (1.21 • N) ( Fig. 3 ; Table 1 ).However, eight species demonstrated latitudinal distribution shifts that were inconsistent with the expected pattern of poleward movement, with cusk, little skate, northern shortfin squid, offshore hake, thorny skate, windowpane flounder, winter skate, and witch flounder all shifting significantly toward more southern latitudes ( Table 1 ).In this study region, latitudinal shifts northward generally correspond with east- ward longitudinal shifts, and vice versa.However, American plaice exhibited a significant westward longitudinal shift with a northward latitudinal change ( Supplementary Fig. S4 ; Table 1 ).Westward shifts can enable species to move toward deeper waters, such as off the shallow plateau of Georges Bank or into the Gulf of Maine's deep basins, and a deepening distribution is also observed for American plaice ( Table 1 ).In total, 23 of 49 species experienced distribution shifts toward deeper depths, while two species-blackbelly rosefish and windowpane flounder-occupied shallower depths in the 2010-2019 period ( Supplementary Fig. S5 ; Table 1 ).Seven species deepened their distribution by > 25 m during the 2010-2019 period, with the strongest deepening observed for red hake (29.8 m), fawn cusk-eel (29.7 m), and alewife (29.0 m).
While the observed shifts in latitude and depth supported the directionality of the distribution change we hypothesized, most species did not all maintain stable thermal environments.Of the 49 species analyzed, 38 experienced warmer surface temperatures and 36 experienced warmer bottom temperatures during 2010-2019 ( Supplementary Figs S6 and S7 ; Table 1 ).Some of the strongest increases in surface temperature were encountered by northern sand lance (2.43 • C) and northern shortfin squid (2.39 • C); the strongest bottom tempera-ture increases were observed for haddock (1.59 • C), offshore hake (1.46 • C), and cusk (1.37 • C).In contrast to the dominant warming patterns, one species (i.e.American lobster) showed a significant decline in bottom temperature, and one (i.e.blackbelly rosefish) experienced a significant decrease in surface temperature during 2010-2019.
Combining the distribution and temperature results reveals that species differ in whether they are able to shift distribution to track temperatures, essentially enabling them to maintain constant temperatures in the habitats they occupy within a warming ocean region.In this study, 9 species shifted their latitudinal, longitudinal, or depth distributions and avoided experiencing surface or bottom warming during 2010-2019, despite the general warming in the region; we term these species "effective trackers" of temperature since their distribution shifts enabled them to maintain constant temperature conditions ( Fig. 4 ).Over 70% of the species we analyzed-35 total-shifted some aspect of their distribution but did not avoid the rapidly warming regional temperatures, a situation that makes them "ineffective trackers" of temperature.Five species appeared to be "non-trackers" in that they experienced warming and did not shift any aspect of their spatial distribution.To summarize, our hypothesis that distribution shifts Only statistically significant differences between the two time periods are reported.
would enable species to maintain stable temperatures in the habitats they occupy was supported for 18% of the species we analyzed (the "effective trackers"), but it was refuted by patterns observed for 82% of the species (the "ineffective trackers" and "non-trackers") ( Fig. 4 ).

Size and growth
Size-at-age analyses enabled us to discern how growth and size have changed over multiple life stages for 14 species ( Fig. 5 , Supplementary Figs S8 , S9 , and S10 ; Supplementary Table S3 ).
Comparing 1970-2009 and 2010-2019, we observed declines in length ( Fig. 5 ; Supplementary Table S3 ) and weight ( Supplementary Fig. S8 ; Supplementary Table S3 ) across all ages examined for four species-haddock, scup, winter flounder, and yellowtail flounder.Six species were observed to have larger lengths and weights at early ages, before showing declines in size at older ages; these species included Ameri-can plaice, Atlantic herring, Atlantic mackerel, pollock, white hake, and witch flounder.Growth of Atlantic cod showed mixed results across these two dominant patterns.Atlantic cod declined in length across all ages but had larger weights at age 1 and smaller weights at older ages during 2010-2019.Silver hake stood out as the only species showing larger length and weight across all ages during the 2010s decade.
The magnitude of changes in size varied widely among species ( Supplementary Table S3 ).When comparing changes in size during the early years of life, the largest length changes were seen in silver hake (age 2, + 11.3%), witch flounder (age 2, + 10.2%), and pollock (age 1, + 8.3%).For the oldest ages that showed significant differences, the largest changes in length were seen in haddock (age 10, −16.9%),American plaice (age 9, −15.7%), and winter flounder (age 8, −14.1%  all increasing by > 30% during the 2010s decade compared to preceding years.However, weight-at-age showed substantial declines for the oldest ages of winter flounder, haddock, American plaice, and yellowtail flounder, all of which experienced weight reductions of > 30% during the 2010s decade. Decadal patterns of size-at-age reveal that, for many of the species examined, size-at-age was smaller across all ages or for older ages during the 2010s than for any of the preceding decades ( Supplementary Figs S11 and S12 ).This pattern is illustrated for length-at-age ( Supplementary Fig. S11 ) and weight-at-age ( Supplementary Fig. S12 ) by haddock, pollock, scup, witch flounder, and yellowtail flounder.For some species, such as witch flounder, the size-at-age of young ages shows a distinct increase over time.However, this decadal distinction was not as clear for other species, which showed the 2010s as intermediate to other decades in terms of weight-at-age and length-at-age.

Population productivity
Changes in productivity as indexed through the spawnerrecruit ratio were also observed during the 2010s decade.For nine of the 15 stocks, productivity was significantly different pre-and post-2010, yet these changes varied in directionality ( Table 2 , Fig. 6 ).Seven stocks showed significantly lower productivity, and for six of these stocks, the declines in productivity exceeded 50% ( Table 2 ).Declines were greatest for scup ( −89%) and winter flounder in the Southern New England-Mid Atlantic stock ( −70%).Two of the stocks that showed significant changes in productivity experienced in-creases in R/SSB.These stocks included pollock, which had a 72% increase, and Atlantic mackerel, which experienced a 200% increase in R/SSB.However, an analysis of potential high-leverage years found that statistically significant results for both of these stocks were influenced by individual years in the data ( Supplement S2 ).
In many of the stocks we examined, changes in SSB, rather than R, strongly influenced the productivity patterns.Recruitment of American plaice, black sea bass, and scup during the 2010s was generally within the range of prior years, and some high R peaks were observed ( Supplementary Fig. S13 ); yet R/SSB was low due to substantial increases in SSB for these stocks.The SSB of American plaice was 66% greater after 2010 than before, black sea bass was 400% larger, and scup was 336% higher ( Supplementary Fig. S14 , Supplementary Table S4 ).In contrast, high productivity of mackerel was associated with extremely low SSB, which declined fairly steadily since 1970 and was at low levels during the 2010s ( Supplementary Fig. S14 , Supplementary Table S4 ).
Many of the stocks that did not have statistically significant changes in R/SSB during the 2010s decade showed modest declines in productivity.However, the Georges Bank and Gulf of Maine haddock stocks both experienced substantial (but not statistically significant) increases in productivity ( Fig. 6 ; Table 2 ), which were driven by recruitment of one strong year class ( Supplementary Fig. S13 ).High variability in R/SSB is seen across many species over the full time period, but the influence of one strong recruitment event during the 2010s decade stands out for the two haddock stocks ( Fig. 6 ).S2 .Species in bold demonstrated changes in size-at-age that conformed with expectations associated with warming; silver hake is underlined to indicate that its size-at-age changes were not consistent with expectations for warming.Species in italics demonstrated significant changes in population productivity.

Reflecting on the 20 1 0s decade
For the Northeast US Shelf, the 2010s decade was characterized by an exceptionally warm ocean temperature regime.During this decade, average annual SST was more than a degree Celsius warmer than temperatures experienced during the preceding three decades, and warming in this region substantially outpaced that experienced in most other parts of the world's oceans.Although we examined warming at the ocean's surface, other studies have demonstrated that warming permeated the water column and affected bottom waters as well (Record et al. 2019, Friedland et al. 2020, Du Pontavice et al. 2023 ).The strong warming trend has been punc-tuated by more frequent and persistent marine heatwaves, as overall warming pushed temperatures further away from historical baselines.The first major heatwave in 2012 led to rapid ecosystem changes and economic disruptions in fisheries (Mills et al. 2013 ).Since then, marine heatwaves have become a recurring event, with heatwaves spanning a portion of each year, particularly during the late summer and fall months.This finding of increasing marine heatwaves is associated with the use of a constant historical baseline (Hobday et al. 2016 )  The warm ocean temperature regime experienced since 2010 was associated with changes in multiple biological processes for many fish and invertebrate species.Species distribution shifts on the Northeast US Shelf have been the subject of a number of prior studies (Nye et al. 2009, Pinsky et al. 2013, Kleisner et al. 2016 ).This paper extends empirical analyses through the end of the 2010s decade, with a particular focus on how distribution metrics differed during the 2010s relative to 1970-2009.In some cases, differences represented continued shifts that had started during previous decades (e.g.red hake depth, Supplementary Fig. S5 ), yet other differences stood out as distinct changes during the 2010s (e.g.yellowtail flounder latitude, Supplementary Fig. S3 ).Differences documented during the 2010s reinforce previous findings of spatial distribution shifts by many fish and invertebrate species that are diverse in terms of taxonomy, life histories, and habitat preferences.In most cases, the distribution shifts were toward more northern latitudes, eastern longitudes, or deeper depths-all of which are consistent with directionalities expected in warming habitats in the region.However, in some cases, "wrong-way" shifts (Fuchs et al. 2020 ) were observed, with eight of the species we analyzed demonstrating southerly changes in latitude and two shifting toward shallower depths.We hypothesized that "wrong-way" shifts may be associated  with increasing population sizes that push species into spaces beyond their preferred habitats (MacCall 1990 ).While blackbelly rosefish followed this pattern of shifting toward shallower depths at higher abundance levels during the 2010s ( Supplementary Figs S5 and S15 ), it was not observed for the other species that shifted opposite of expected directions in terms of latitude and depth.Similarly, prior studies in the region have reported variability in relationships between species distribution changes and population abundance (Murawski and Finn 1988, Nye et al. 2009, Kleisner et al. 2016 ).
Unlike previous studies, our results indicate that species distribution shifts did not enable most species to maintain stable temperatures within their occupied habitats during the 2010-2019 period.Studies of distribution changes through the late 2000s or early 2010s indicated that species were able to track climate velocities and remain within their preferred temperature ranges by shifting their distributions (Nye et al. 2011, Pinsky et al. 2013 ).We found that most species were not able to keep pace with the rapid warming experienced in surface and bottom temperatures during the 2010s by shifting their distributions.In this study, 71% of the species examined proved to be ineffective at tracking the pace of warming in the region; they demonstrated a distribution change (i.e.latitudinal, longitudinal, or depth) but still experienced warmer surface and/or bottom temperatures.Only 18% of the species were able to avoid warmer temperatures as they shifted distribution.
In ectotherms, a key biological process that may be affected by warming is growth.According to the temperature-size rule, fish in warmer waters are expected to grow faster during early life stages but plateau at smaller adult sizes (Atkinson 1994 ).However, under extreme thermal stress, growth may be curtailed in early life stages as well (e.g.Guo et al. 2022 ).For the species we examined, we saw a preponderance of growth patterns that aligned with expectations for warmer ocean temperatures, specifically: (1) larger size at young ages but smaller sizes at older ages; and (2) declining size across all ages.Twelve of the 14 species (86%) we analyzed demonstrated one of these two patterns in length-and weight-at-age during the warm 2010s decade compared to preceding decades ( Fig. 5 , Supplementary Fig. S8 ).Only silver hake demonstrated faster growth and larger sizes across all ages during the 2010-2019 period, while results were inconsistent for butterfish.
In addition to alignment between warming expectations and observed growth patterns, the magnitude of differences in size-at-age during the 2010s compared to prior decades was substantial.At the maximum age we evaluated for each species, length declined between 1.2% and 17%, while weights declined between 6% and 37% for most species, aside from silver hake (increased) and butterfish (no significant difference) ( Supplementary Table S3 ).Half of the 12 species showing significant declines in weight-at-age were 20% smaller in the 2010s at the maximum age analyzed.For species experiencing the strongest declines in length, differences observed between the 2010s and prior decades on the Northeast Shelf were within the range observed over more than four decades in the North Sea (Baudron et al. 2014 ) S10 ).In one instance, smaller size at all ages of a strong year class (2003) of Georges Bank haddock has been associated with density dependence (Brodziak et al. 2008 ).Density dependence likely also constrained the growth of haddock during the 2010s due to strong year classes in 2010 and 2013 (Leaf and Friedland 2014 ).In another case, age-2 cod on Georges Bank showed declines in size by 2000, yet age-2 cod in the Gulf of Maine were larger after 2010 than previously (McBride et al. 2022 ).For these two stocks of cod, subregional differences counterbalance one another in our shelfwide analysis.
Declines in individual growth can curtail the productivity of populations, as fecundity scales with body mass and large female spawners have additional reproductive advantages that enhance their contributions to population growth relative to smaller spawners (Hixon et al. 2014, Barneche et al. 2018 ).In the Gulf of Maine, warming-related declines in weight-atage of Atlantic cod, particularly around the age at maturity, have been associated with declines in stock productivity (Pershing et al. 2016 ).Our current analysis does not draw a direct link between temperature, growth, and stock productivity, but it does document declines in the population productivity of a number of stocks in the Northeast Shelf region during the warm 2010s decade, when size-at-age also declined for many species.These findings reinforce those of other studies that demonstrated time-varying productivity (Tableau et al. 2019 ) and a sharp decline in recruitment for many species during the early 2000s (Perretti et al. 2017 ).For some stocks, productivity during the 2010s decade was comparable to and even lower than during the low-recruitment regime of the 2000s.
warm temperature regime.During the 2010s decade, warm ocean temperatures stood out against prior observations, and marine heatwaves became a regular occurrence.Most species could not buffer themselves against this rapid increase in temperatures by shifting their distributions, and effects consistent with living under warmer temperatures were apparent in growth patterns for most fish species.In addition, productivity declines were observed in some populations, all of which had experienced reduced individual growth.Although we did not explicitly examine linkages between and mechanisms driving these processes, their align-ment with the warm temperature regime indicates that climate change is impacting fish populations and the Northeast Shelf marine ecosystem through a variety of direct and indirect pathways.

Looking ahead
Temperatures that occurred on the Northeast Shelf during the 2010s placed the region at the forefront of the world's oceans in terms of warming rates, and marine heatwaves put temperatures on par with those projected by the end of the century under certain climate scenarios (Mills et al. 2013 ) projections indicate ocean waters in this region will continue to warm at exceptional rates in the future (Saba et al. 2016, Alexander et al. 2020 ).Moreover, the frequency and duration of marine heatwaves will continue to increase as temperatures move further away from those defined during historical climatological periods (Hobday et al. 2016, Alexander et al. 2018, Brickman et al. 2021 ).Studies that have extended projections of temperature to biological processes also show ongoing changes in fish populations and communities.
Projections of species distributions indicate future increases of suitable thermal habitat for species currently in the Mid-Atlantic Bight sub-region and declines for species in the Gulf of Maine/Georges Bank sub-regions, resulting in northward shifts of southern species and declines in northern species on the Northeast Shelf (Kleisner et al. 2017, Allyn et al. 2020 ).Similar findings have resulted from population projections under warming conditions, with higher temperatures causing declines of stocks at the upper end of their range of thermal tolerance, while benefiting those that are not near this limit (Pershing et al. 2015, Le Bris et al. 2018 ).
Given the population and ecosystem changes that are projected in the coming decades, the experience of the 2010s warm temperature regime in the Northeast Shelf region provides an opportunity to build a greater understanding of how increasing (warming trends), persistently high (warm regimes), and intermittently extreme (heatwaves) temperatures all affect fish populations.Using these experiences as a basis for further research will be important for advancing predictive models of potential future climate impacts.First, research is needed to understand the direct and indirect mechanisms through which warming affects fish populations.Direct physiological or behavioral mechanisms associated with thermal preferences and tolerances may cause species to shift distributions or affect metabolism and growth.However, temperature can also indirectly influence each process through a variety of ecosystem changes, such as altered prey abundance and quality (Pershing and Kemberling 2023 ), predation rates (Friedland et al. 2012 , Richards andHunter 2021 ), or fishing interactions (Pershing et al. 2015 ).Secondly, while this study focused on the 2010s decade, some changes began in earlier years and may not have been initiated by warming.Integrated analyses of the influence of additional drivers (e.g.density dependence, fishing pressure) and their interactions with warming are needed to determine the mechanisms associated with changes in population processes and understand the influence of climate change within the context of other stressors and conditions.Finally, research to establish linkages between biological processes is necessary for elucidating how warming effects permeate through populations and ecosystems.For example, productivity changes may shape distribution patterns, growth changes may alter predation rates, and these and other linkages may influence population and community outcomes.It is likely that linkages between biological processes take multiple forms depending on species-specific responses to warming and other ecosystem conditions.These topics merit detailed analysis in future studies.
Although this paper has focused on the physical and biological characteristics of the Northeast Shelf during a warm regime, challenges to fisheries and fishery management are also arising (Mills et al. 2013, Pershing et al. 2015 ).Ongoing warming in the region will require management and decisionmaking processes to proceed in the context of trends and shifts that push ecosystem conditions beyond past analogues and that create higher levels of uncertainty (Pershing et al. 2019, Bell et al. 2020 ).Considerations of time-varying productivity, dynamic reference points, and rebuilding timelines are expanding as fishery scientists and managers strive to align stock advice, harvest levels, and conservation measures with changing ecosystem conditions (Pershing et al. 2015, Britten et al. 2017, Tableau et al. 2019, Holsman et al. 2020 ).Ecosystem regimes and non-stationary relationships are increasingly being considered in stock assessment models and advice processes (e.g.Miller et al. 2016 ), yet forecasting the persistence or erosion of states such as the warm temperature regime studied here remains a nascent endeavor (Hunsicker et al. 2022 ).While forecasting efforts advance, management approaches that support decision-making under uncertainty will be increasingly important for sustaining fisheries in the context of warming trends, temperature regimes, and heatwave events associated with climate change.
M.B. conducted the data analysis, and all authors contributed to the interpretation of results.K.E.M. led the writing of the manuscript, and all authors contributed to writing and editing and approved the final draft.

Figure 1 .
Figure 1.Map of the study area, with surv e y strata used in this analysis outlined with thin black lines.K e y geographic regions referred to in the paper are shaded in distinct colors.Stratum numbers are given in Supplementary Fig. S1.1 .
SST on the Northeast US Shelf has been warming at a rate of 0.38 • C per decade since 1982, a rate that is 2.5 times faster than the warming rate experienced over the global oceans during that time (0.15 • C/decade).The average SST in the most recent decade (2010-2021) was 12.93 • C, representing an average anomaly of 1.06 • C; for the decades spanning 1982-2009, the annual SST averaged 11.83 • C, representing an anomaly of −0.043 • C ( Fig. 2 a).The strong positive SST anomalies since 2010, with annual anomalies consistently above 0.5 • C and reaching a peak of 1.92 • C in 2012, indicate a temperature regime shift in the region.The difference in the strength and consistency of warm SST anomalies confirms the distinctness of the 2010s decade in terms of temperature.

Figure 2 .
Figure 2. Sea surface temperature on the Northeast Shelf.(a) Annual average SST anomalies (dots) and means (solid thick lines) for the time periods 1982-2009 (blue) and 20 1 0-2021 (orange).(b) Map of annual warming rates (1982-2021) o v er the region.(c) A heatmap of the daily temperature anomalies from 1 January 1982 to 31 December 2021, with days that qualify as marine heatwaves indicated with black horizontal lines.

Figure 3 .
Figure 3. Map of distribution shifts for species that showed significant changes in latitude and longitude during 20 1 0-20 19 compared to 1970-2009.Blue points indicate the center of biomass of the species during 1970-2009, and orange points indicate the center of biomass for 20 1 0-20 19.

Figure 4 .
Figure 4. Classification of species based on their spatial distribution changes and resulting temperature exposure during the 20 1 0s decade relative to 1970-2009.Species were classified as shifting distribution if they demonstrated a significant change in latit ude, longit ude, or depth; they were classified as experiencing warming if they demonstrated an increase in either surface or bottom temperature.Maps showing examples of distribution shifts associated with each category are a v ailable in Supplementary Fig.S2.Species in bold demonstrated changes in size-at-age that conformed with expectations associated with warming; silver hake is underlined to indicate that its size-at-age changes were not consistent with expectations for warming.Species in italics demonstrated significant changes in population productivity.
, and additional methods to identify true temperature "shocks" in the context of a warming ecoregion are being explored (e.g.Jacox et al. 2020 , Amaya et al. 2023 ).

Figure 5 .
Figure 5. Percent difference in length-at-age for 1970-2009 versus 20 1 0-20 19.Blue bars indicate significant increases in length, while yellow bars indicate significant declines.Gray bars denote ages for which there were not significant differences in length.

Figure 6 .
Figure 6.Time series of the R-to-SSB ratio for 15 stocks, indicating the mean ratio prior to 20 1 0 in blue and from 20 1 0 on w ards in orange.Asterisks f ollo wing the stock name indicate a statistically significant difference in the ratio during the two time periods.Stock names are fully described in Supplementary TableS2.

Table 2 .
Means of the recruits (R) per spawner (SSB) ratio for 15 stocks during years prior to 20 1 0 and for 20 1 0 onwards.Statistically significant differences in the R/SSB ratio for the two time periods are identified in bold.Stock abbreviations are as follows: CCGOM = Cape Cod-Gulf of Maine; GB = Georges Bank; GOM = Gulf of Maine; SNEMA = Southern New England-Mid Atlantic.
. Weight declines were on par with and even exceeded modelbased projections for warming between 2000 and 2050 (Cheung et al. 2013 a).The empirical results for the Northeast Shelf provide an additional demonstration ( sensu Baudron et al. 2014 ) that substantial declines in body size can occur rapidly under certain conditions.While this analysis focused on comparing size-at-age during the 2010s versus 1970-2009, changes in size and growth of certain species began prior to 2010 ( Supplementary Figs S9 and Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsad208/7596727 by guest on 06 February 2024 Multispecies population-scale emergence of climate change signals in an ocean warming hotspot 11 . Climate Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsad208/7596727 by guest on 06 February 2024