Redefining North Atlantic right whale habitat‐use patterns under climate change

Changes in the physical oceanography of the Northwest Atlantic stemming from both natural and anthropogenic climate change impact the foraging ecology and distribution of endangered North Atlantic right whales. In this study, right whale sightings from 1990 to 2018 were analyzed to examine decadal patterns in monthly habitat use in 12 high‐use areas. Depth‐integrated abundances of late‐stage Calanus finmarchicus and Calanus hyperboreus were also analyzed for decadal variations in the right whale foraging habitats. There were significant differences in the occurrence, seasonal timing, and persistence of foraging habitats across these three decades. In the decades of the 1990s and the 2010s, prey was less abundant than in the 2000s, corresponding to reduced use of the Southeast US calving grounds in the winter, increased use of Cape Cod Bay in winter and spring, and reduced use of Roseway Basin in the fall. In the 2010s, right whale sightings increased in Southern New England and the Gulf of St. Lawrence in the spring and summer, respectively. Summertime declines in the 2010s in late‐stage copepod abundances in the Gulf of Maine and surrounding regions, as well as in the Gulf of St. Lawrence, indicate that recent increased use of the Gulf of St. Lawrence is driven by a decline in prey in traditional foraging habitats rather than by an increase in prey in the new foraging habitat. This analysis of decadal‐scale differences in right whale sightings and prey abundance is critical for redefining right whale distribution patterns for the most recent (post‐2010) decade.

Warming in the Northwest Atlantic is occurring faster than most of the global ocean Seidov et al. 2021). Climate-driven patterns in warming and ocean circulation in this region have been associated with declines in calanoid copepods (Record et al. 2019;Sorochan et al. 2019;Meyer-Gutbrod et al. 2021), and shifts in the abundance and distribution of higher trophic levels (Nye et al. 2009;Pinsky et al. 2013). These oceanographic processes have altered the foraging environment and behavior of the critically endangered North Atlantic right whale (Eubalaena glacialis), reducing the population's calving rate and exposing it to greater risks of serious injury and mortality from ship strikes and entanglement in fishing gear (Meyer-Gutbrod et al. 2021). With reduced reproduction and elevated mortality rates, the right whale population has exhibited a multi-year decline for the first time since post-whaling demographic data became available during the early 1980s. These findings illustrate an important indirect effect of climate change. As climate-driven changes in prey availability alter the habitat use of marine species, predators may shift their foraging behaviors and locations and become exposed to elevated rates of anthropogenic mortality.
One of the major obstacles to the successful implementation of vessel and fishery management is the uncertainty in right whale distribution and seasonal habitat use. Right whales migrate along the continental shelf of the western North Atlantic, with typical habitats ranging between Florida, US and Newfoundland, Canada, although sporadic detections have occurred outside of this range (Knowlton et al. 1992; Kraus and Rolland 2007;Davis et al. 2017). Conservation measures have previously been focused in areas and during seasons when right whale densities are high. However, predicting right whale distribution, especially at the high spatial and temporal scales relevant for policy implementation, is challenging (Ross et al. 2021). Right whales have high daily nutritional requirements and derive their nutrition from zooplankton (Fortune et al. 2013), particularly Calanus finmarchicus during the lipidrich late copepodite life stages (Lee et al. 2006). Understanding spatial and temporal shifts in high-density prey aggregations is essential for predicting right whale distributions.
During the period of 1980 through 2010, the annual right whale distribution cycle was conceptualized as a seasonal pattern driven by the separation in space of calving and foraging, with some regional interannual variability (e.g., Winn et al. 1986). In this conceptual model, pregnant females travel to the shallow waters off the southeastern United States to give birth in the winter months, typically between December and April (Hamilton and Cooper 2010;Keller et al. 2012;Soldevilla et al. 2014). They are often accompanied by juvenile right whales and occasionally, non-reproductive females and adult males (Kraus et al. 1986;Gowan et al. 2019). Prey is not thought to be available along the migration route across the Mid-Atlantic Bight or in the Southeast US calving ground, therefore right whales fast during the calving ground migration (Fortune et al. 2013). In the winter and spring, right whales typically converge in high densities in and around Cape Cod Bay where they feed on Pseudocalanus and Centropages spp. in winter and C. finmarchicus in spring (Mayo and Marx 1990;DeLorenzo Costa et al. 2006). As spring progresses, right whales emerge from Cape Cod Bay and feed on late-stage C. finmarchicus in the Great South Channel (Kenney et al. 1995). In the summer and fall, high densities of right whales have historically been found in the Grand Manan Basin, Bay of Fundy, and Roseway Basin on the western Scotian Shelf foraging on diapausing, lipid-rich C. finmarchicus (Baumgartner et al. 2003;Michaud and Taggart 2011;. Divergences from typical seasonal habitat-use patterns have been observed in response to interannual variation in prey availability driven by climate forcing. Abundance of C. finmarchicus is sensitive to trophic and demographic impacts at the regional scale (Frank et al. 2005;Ji et al. 2022), fluctuations in temperature and salinity driven by shifting water masses (MERCINA 2012;Davies et al. 2014;Meyer-Gutbrod et al. 2021) and alterations in advective patterns that impact C. finmarchicus downstream supply (MERCINA 2004;Runge et al. 2015;Ji et al. 2017). Decadal-scale regime shifts in the oceanography of the Northwest Atlantic over the past 40 years explain considerable variation in late-stage C. finmarchicus abundance in the Gulf of Maine region (Greene et al. 2013), and have been shown to directly impact right whale demography and distribution Meyer-Gutbrod et al. 2021). In 1992, right whales passed over their typical spring foraging in the Great South Channel, likely in response to a shift in the zooplankton community which occurred in conjunction with anomalously low salinity (Kenney 2001). From 1993 to 1997, right whales abandoned their use of Roseway Basin on the western Scotian Shelf, again in response to low prey densities (Patrician and Kenney 2010;Davies et al. 2015). Due to the high energetic demand of reproduction (Fortune et al. 2013), these periods of low prey availability directly impacted population growth due to decreased calving rates (Greene and Pershing 2004;Meyer-Gutbrod et al. 2015).
Starting in 2010, a major shift in the annual cycle of right whale occurrence appears to have taken place in response to an oceanographic regime shift which warrants a reevaluation of right whale seasonal habitat-use patterns. Since 2010, right whale use of the Gulf of Maine, Bay of Fundy, and Scotian Shelf regions has declined substantially (Davis et al. 2017;Record et al. 2019), and in the latter half of the decade, a large portion of the population has begun to utilize the southern Gulf of St. Lawrence during spring, summer and autumn (Simard et al. 2019;Crowe et al. 2021). This emerging high-use habitat may be an important area for foraging on latestage C. finmarchicus, as well as Calanus hyperboreus Lehoux et al. 2020;Brennan et al. 2021). This shift in distribution resulted in mortalities beginning in 2015 in the Gulf of St. Lawrence, with an unexpected mortality event declared in 2017, as right whales became more vulnerable to ship strikes and entanglements in unprotected waters Davies and Brillant 2019). A concurrent decline in reproduction may have occurred due to foraging gaps or relatively low prey in the Gulf of St. Lawrence (Gavrilchuk et al. 2021), however, recent calving rates are higher among females that use the Gulf of St. Lawrence (Bishop et al. 2022).
The right whale range contains at least a dozen known areas where there is a relatively high frequency of detections. These high-use areas include regions important for foraging, breeding, and transiting, and more are discovered as survey effort expands to previously under-surveyed areas. Long-term variability in right whale occurrence has been documented in many of these high-use areas (e.g., Mayo et al. 2018;Charif et al. 2020). However, there is a need to characterize the historical and contemporary variation in right whale distribution within and among all high-use areas, which are connected contiguously along the migratory route. Previous studies have used regime shift detection analysis to reveal distinct oceanographic regimes in the 1990s, 2000s, and 2010s that govern the spatial and temporal patterns of C. finmarchicus abundance and, in turn, right whale reproduction (Greene et al. 2013;Meyer-Gutbrod et al. 2021). Comparisons of fine-scale right whale seasonal habitat use and prey availability between each of these three decades will provide a more comprehensive view of right whale distribution, including the impacts of climate change.
In this study, right whale sightings from 1990 to 2018 were analyzed to examine decadal patterns in monthly habitat use in 12 high-use areas spanning the Gulf of Maine and surrounding waters, Scotian Shelf and Gulf of St. Lawrence, and the calving ground in the Southeast United States. Changes in historical habitat-use patterns were characterized between the decades of the 1990s, 2000s, and 2010s, and emerging habitat-use patterns were identified. Water column-integrated abundances of late-stage C. finmarchicus and C. hyperboreus were also analyzed for decadal variations within the high-use right whale foraging habitats or upstream of these habitats. Decadal variations in both right whale and prey abundances in each habitat were assessed to identify coherent patterns in predator and prey. By comparing the historical distributions with contemporary patterns, we redefined right whale annual distribution patterns for the post-2010 decade.

Right whale sightings per unit effort
The North Atlantic Right Whale Consortium maintains a Sightings Database containing more than 56,000 right whale sightings and associated survey effort collected during shipboard and aerial surveys (North Atlantic Right Whale Consortium 2020a, 2020b). Survey data are collected from relatively broad areas of the Northwest Atlantic, but survey effort is concentrated in high-use right whale habitats, including the Southeast US calving ground off the coast of Georgia and Florida, the Gulf of Maine, the Scotian Shelf, the Bay of Fundy and, recently, the Gulf of St. Lawrence (Brown et al. 2007). Sightings and survey effort compiled by the Consortium were used to create a spatial grid of sightings and a corresponding grid for effort. Sightings Per Unit Effort (SPUE) values were calculated over a 10-Â 10-min raster grid for each month in each year over the time period 1990-2018. Sightings consist of any positive or probable identification of the North Atlantic right whale species, but not necessarily an identification of the individual right whale. Effort was defined as the number of kilometers of track line surveyed at a Beaufort sea state between 0 and 4, with at least 3.7 km (2 nautical miles) visibility, with at least one observer recorded as on watch, and (for aerial surveys) altitude below 366 m (1200 ft). SPUE was calculated as the total number of right whales sighted divided by the total effort within a given grid cell in a given month and year.
Additional sightings and effort data were collected by aerial platforms commissioned by Fisheries and Oceans Canada and Transport Canada during 2017 in the Gulf of St Lawrence. These data are not included in the Consortium Sightings Database at this time, and were contributed separately by the organizations. These sightings and effort data were processed and gridded in the same manner as above and incorporated into the right whale SPUE calculations.
Due to the paucity of survey effort in the Gulf of St Lawrence prior to the Unusual Mortality Event in 2017 (Supporting Information Table S1), we used Mingan Island Cetacean Study sightings and survey effort from 1980 through 2019 to assess the possibility of right whale presence in the Gulf of St. Lawrence in previous decades. Surveys are conducted on small coastal vessels, primarily targeting rorqual whale sightings and identification; however, data are collected for all observed cetaceans (Ramp and Sears 2013). For this study, we examined North Atlantic right whale sightings from surveys run in the spatial domain of 49.5-50 N and 62-64.5 W, which includes the Jacques Cartier Passage between Mingan Island and Anticosti Island, and adjacent waters (Fig. 1). Survey effort was quantified as the number of survey days per year, and effort was concentrated in the months of June through October. Daily sightings included the first sighting of each unique right whale individual by each vessel crew, with between one and four vessels operating on a given day. Although the Mingan Island Cetacean Study effort comprises a first approximation, the high number of effort days over each summer covering a small spatial region indicates that most right whales in the region had a high probability of being observed. Due to the difference in effort records, these data were not incorporated into the right whale SPUE analysis, but they are an important source of information on variation in right whale use of the Gulf of St. Lawrence over the past four decades.

Zooplankton data
Data from the NOAA Northeast Fisheries Science Center Ecosystem Monitoring program (EcoMon), and the Fisheries and Oceans Canada Atlantic Zone Monitoring Program (AZMP) were used to examine decadal regional-scale variation in prey abundances to assess their potential role in driving variation in right whale habitat use.
For the EcoMon program, net tows were conducted biweekly, monthly or quarterly, depending on the year, throughout the Gulf of Maine and along the New England shelf (Kane 2007;Hare 2021). In this study, samples were binned quarterly over the time series 1990-2019. Plankton samples were collected both day and night with 0.61-m diameter bongo nets fitted with a 333-μm mesh net. Oblique tows were conducted for a minimum duration of 5 min and spanned the water column to a depth within 5 m of the seafloor, or to a maximum depth of 200 m. A flowmeter attached to the nets measured the volume of seawater filtered. Zooplankton were preserved, counted, and identified to species and life stage according to a standard protocol (Hare 2021). Zooplankton abundances were normalized to the count of organisms beneath 10 m 2 of sea surface using standard haul factors. A total of 11,153 net tows were analyzed across nine right whale habitat polygons, and sampling frequency within each habitat polygon was generally proportional to the polygon area.
Net tow data from the AZMP (Therriault et al. 1998 (Fig. 1). Zooplankton tows were conducted using 0.75-m diameter ring nets fitted with 202-μm mesh. Nets were towed vertically from near bottom to the surface. Zooplankton were preserved, counted, and identified to species and life stage according to a standard protocol (Mitchell et al. 2002). Zooplankton abundances were normalized to the count of organisms beneath 1 m 2 of sea surface.
Quarterly zooplankton abundance values were calculated separately for both the EcoMon and the AZMP programs. Late-stage C. finmarchicus abundances were calculated by aggregating the associated C. finmarchicus copepodite stages C5 and C6 abundances. With the AZMP data, C. hyperboreus copepodite stages C4, C5, and C6 were aggregated to create late-stage C. hyperboreus abundance values at the four AZMP stations. EcoMon zooplankton abundances were aggregated across each distinct polygon corresponding to right whale habitats ( Fig. 1, Table 1). AZMP zooplankton abundances were presented separately for each of the four stations included in the analysis.

Spatial analysis
Monthly SPUE grid cells and zooplankton sampling locations were separated into broad-scale spatial polygons defined  Table 1. Right whale SPUE values in the 10-Â 10-min grid cells were associated with the polygon that contains the center of the grid cell.

Statistical analysis
We used generalized linear models to test the alternative hypotheses that right whale occurrence and prey abundance differed between decades. In these models, we initially used a Poisson distribution, which is appropriate for count data, and tested for overdispersion using the AER package (Kleiber and Zeileis 2008; R Core Team 2019). When overdispersion was found, a quasi-Poisson distribution was run in place of a Poisson regression.
To test the hypothesis that right whale habitat use has changed among three decades from 1990 to 2018, separate models were developed for each combination of habitat polygon and month to compare right whale sightings between consecutive decades. The number of right whale sightings for a given month and polygon was modeled as follows: where N is the number of right whales sighted within a specific habitat polygon, h, and month, m. Decade, D, is a binomial variable representing two consecutive decades, and effort, E, within the specific habitat polygon, h, and month, m, was included as an offset term. This set of models was run once to compare sightings between the 1990s and the 2000s, and another set of models was run to compare sightings between the 2000s and the 2010s. Decadal comparisons were made relative to the 2000s when right whale population growth was the highest (Pace et al. 2017), so positive differences in SPUE reflect more whales in the 2000s, and negative differences in SPUE reflect more whales in either the 1990s or the 2010s. This comparison of the 1990s and 2010s relative to the 2000s was useful for identifying similar habitat-use strategies between the 1990s and 2010s. Models were only run when at least one sighting occurred for a given combination of habitat and month. A similar set of models was developed to compare the effect of decade on C. finmarchicus abundance using the EcoMon surveys. The models were formulated as where C is the abundance of late-stage C. finmarchicus for a given habitat polygon, h, and quarter, q. There are four quarters per year: Quarter 1 = January-March, Quarter 2 = April-June, Quarter 3 = July-September, and Quarter 4 = October-December. Decade, D, is a binomial variable representing two consecutive decades. Similar to the right whale sightings models, a set of models was run to compare C. finmarchicus abundances between the 1990s and the 2000s, and another set was run to compare C. finmarchicus abundances between the 2000s and the 2010s. T is the maximum tow depth. Since data are normalized by counts per 10-m 2 area of ocean surface, values have already been corrected for the volume filtered by the net and no offset effort term is required.
Finally, a distinct model was developed for each of the four AZMP stations, and for each copepod species. The models were formulated as where C is the abundance of late-stage copepods for a given taxon, t (C. finmarchicus or C. hyperboreus), AZMP station, s, and quarter, q. Decade, D, is a binomial variable representing two consecutive decades, but in the case of the AZMP program, only the 2000s and 2010s decades were available. T is the tow start depth.

Changes in habitat use in the traditional range
The SPUE analysis from 1990-2018 across all 12 right whale habitat polygons contained 55,064 sightings of right whales over a total of 4,499,121 km of track line. Survey effort was highly variable among decades and polygons (Supporting Information Table S1). A total of 39 generalized linear models were run comparing right whale habitat use between the 1990s and the 2000s, corresponding to each combination of Table 1. Right whale habitat polygon acronyms. The spatial area corresponding to each habitat is shown in Fig. 1.

Region
Habitat polygon acronyms  Information Table S2). There was a significant effect of decade in 23 of these models ( Fig. 2A), although the effect size of decade varied widely (Supporting Information Table S2). During the calving season (December to March), right whale SPUE was significantly higher in the Southeast US calving grounds in the 2000s, whereas in the 1990s, SPUE was higher in Cape Cod Bay and Massachusetts Bay ( Fig. 2A). In the spring and early summer, right whale sightings were significantly higher in the Great South Channel (May, June) and the Gulf of Maine (May-July) during the 2000s, whereas sightings were significantly higher in the Bay of Fundy in June and July in the 1990s ( Fig. 2A). In August and September, SPUE was significantly higher in the Gulf of Maine in the 1990s, whereas in the 2000s, SPUE was significantly higher in Roseway Basin ( Fig. 2A). Maps depicting SPUE summed over the years within a decade show the broad spatial trends in right whale habitat use between subsequent time periods, and demonstrate patterns in decadal habitat use when survey effort is available but a lack of sightings precludes model comparison. Raster maps aggregating SPUE across Quarter 3 show the extent of survey polygon where right whale sightings occurred in both decades. Box color corresponds to the difference in predicted SPUE (number of whales per 100 km) for a given month and habitat polygon. Black borders around a box indicate that decade was a significant predictor of SPUE. Positive numbers (blue boxes) correspond to SPUE that was higher in the 2000s, and negative numbers (red boxes) correspond to SPUE that was higher in the (A) 1990s and (B) 2010s. The color bar limits were set to 6 and À6 sightings per 100 km, and dark red or dark blue squares may correspond to model predictions that exceed those values (see Tables S2, S3). Refer to Fig. 1 and Table 1 for habitat polygon acronyms. effort and the changes in summer and fall foraging patterns in the 1990s and 2000s. In Quarter 3, right whales were heavily concentrated in the Bay of Fundy in the 1990s (Fig. 3A). Subsequently, in the 2000s, they spread out and occupied a much broader area across the Gulf of Maine and the Scotian Shelf regions over the late summer and fall months (Fig. 3B).
Quarterly SPUE maps for each of the three decadal periods are provided in Supporting Information Figs. S1-S12.
Right whale habitat use was also compared between the decade of the 2000s and the 2010s. A total of 59 models were run, corresponding to each combination of habitat polygon and month where right whales were sighted during both of   Fig. 1 and Table 1 for acronym definitions). these decades (Supporting Information Table S3). There was a significant effect of decade in 31 of these models (Fig. 2B). SPUE declined significantly from November to March in the Southeast US calving grounds in the 2010s relative to the 2000s. Concurrently, SPUE in the winter-spring foraging habitats Cape Cod Bay and Massachusetts Bay increased in the 2010s compared to the 2000s (Fig. 2B). In the 2000s, right whales would exit the Cape Cod Bay and Massachusetts Bay region to forage in the Great South Channel in March-May; however, in the 2010s, SPUE remained high in the Cape Cod Bay and Massachusetts Bay for longer into the spring relative to the 2000s; SPUE was also higher in Southern New England in the 2010s in March and May (Fig. 2B). Changes in SPUE in Southern New England could not be resolved for other winter and spring months; although there was effort during both decades, no sightings were made in the 2000s. Significantly lower SPUE were observed in multiple traditional summer and fall habitats in the 2010s compared to the 2000s. For example, SPUE was significantly lower in the 2010s in the Bay of Fundy in August, September, and November, in the Gulf of Maine in October and December, and in Roseway Basin in September (Fig. 2B).
The spatial extent of survey effort and right whale sightings in Quarter 2 (April-June) in the 2000s and 2010s is illustrated in Fig. 4. Right whale observations increased in the northern portion of the Southern New England habitat in the 2010s, driven primarily by sightings in the month of April. Survey effort across much of this habitat in the 2000s yielded only one right whale observation in 2005, which underscores the shift in the use of this area in the 2010s. Right whales have been observed in the Southern New England habitat each month between November and May during a portion of the 2010s. The 2010s were also characterized by Quarter 2 increases in the use of Cape Cod Bay and Massachusetts Bay, as well as occupation of the Gulf of St. Lawrence (Fig. 4).
By examining model results across the three decades, it is evident that there are similarities in right whale habitat-use patterns in the 2010s and 1990s. Compared to the 2000s, the decades of the 1990s and 2010s both had lower SPUE in the Southeast US during the winter migration period. The corresponding increase in SPUE in Cape Cod Bay and Massachusetts Bay in the 1990s and 2010s relative to the 2000s indicates that a portion of the population was utilizing these habitats in the winter and spring as an alternative to the migration to the Southeast US calving ground. However, the high occupancy of Cape Cod Bay and Massachusetts Bay persisted later into the spring in the 2010s compared to either the 1990s or the 2000s. Right whales utilized the Roseway Basin foraging area less in September in both the 1990s and 2010s compared to the 2000s. However, patterns diverge in the occupation of the Bay of Fundy in the summer. Right whale densities were highest in the Bay of Fundy in June-August in the 1990s and lowest in the 2010s, indicating a consistent decline in the use of this habitat over the three decades (Fig. 2).

Right whales in the Gulf of St. Lawrence
While Mingan Island Cetacean Study data could not be included in the decadal SPUE analysis, the time series of consistently high survey effort in the Jacques Cartier Passage from 1980-2019 provides a complementary view of right whale occurrence in the Gulf of St. Lawrence. Survey effort averaged 50 d yr À1 (Fig. 5C). For most of the time series, no right whales were observed in a given year. However, right whale observations began to occur regularly in 2014, and at least 65 right whale sightings occurred in the Mingan Island Cetacean Study survey area each year between 2016 and 2019 (Fig. 5B). Positive photo identifications confirm that there were 27 unique individuals in this area in 2016 (Fig. 5A).

Decadal-scale changes in prey abundance
Zooplankton sampling effort in the EcoMon program was generally proportional to polygon size (Supporting Information Fig. S13), and there were effort gaps in both EcoMon and AZMP surveys in winter due to the presence of seasonal sea ice in the GSL and poor weather in all areas (Supporting Information Figs. S13, S14). For each pair of decades analyzed with the Eco-Mon data, 35 models were run to compare zooplankton abundances. For the 1990s to 2000s comparison, 20 models had a significant effect of decade and all of these models demonstrated an increase in late-stage C. finmarchicus abundance in the 2000s relative to the 1990s (Fig. 6A). Among significant models, the predicted difference in late-stage C. finmarchicus abundance ranged from 1920 more organisms per 10 m 2 in the 2000s (Georges Bank, Quarter 4) to 364,180 more organisms per 10 m 2 in the 2000s (Bay of Fundy, Quarter 3) (Fig. 6A). Model coefficients and model-predicted C. finmarchicus abundance values for the 1990s and 2000s for each right whale habitat polygon and quarter are provided in Supporting Information Table S4.
For the 2000s to 2010s comparison, 21 out of 35 models had a significant effect of decade, with 18 combinations of habitat polygon and quarter demonstrating higher late-stage C. finmarchicus abundance in the 2000s and three combinations of habitat polygon and quarter demonstrating higher late-stage C. finmarchicus abundance in the 2010s (Fig. 6B).
The three cases where C. finmarchicus abundance was significantly higher in the 2010s all occurred during Quarter 2 (April-June) in the Jeffreys Ledge, Cape Cod Bay, and Great South Channel polygons. Among significant models, the predicted difference in late-stage C. finmarchicus water column abundance ranged from 120,000 more organisms per 10 m 2 in the 2010s (Jeffreys Ledge, Quarter 2) to 346,000 more organisms per 10 m 2 in the 2000s (Bay of Fundy, Quarter 3) (Fig. 6B). Model coefficients and model-predicted C. finmarchicus abundance values for the 2000s and 2010s for each habitat polygon and quarter are provided in Supporting Information Table S5.
Analysis of late-stage C. finmarchicus abundance between the 2000s and the 2010s within Canadian waters using the AZMP data resulted in a total of 14 models across the four sampling stations. Nine of the models had a significant effect of decade and all of these demonstrated a decrease in latestage C. finmarchicus abundance in the 2010s relative to the 2000s (Fig. 7A). Among significant models, the predicted difference in late-stage C. finmarchicus abundance ranged from 560 more organisms per m 2 in the 2000s (Shediac Valley, Quarter 2) to 10,000 more organisms per m 2 in the 2000s (Rimouski, Quarter 4) (Fig. 7A). Model coefficients and modelpredicted C. finmarchicus abundance values for the 2000s and 2010s for each station and quarter are provided in Supporting Information Table S6. Fig. 6. Difference in model-predicted late-stage C. finmarchicus abundance per 10 m 2 between subsequent decades for each right whale habitat and quarter. Box colors indicate differences in predicted abundance between (A) the 2000s and the 1990s, and (B) the 2000s and 2010s, for the median plankton tow depth in the corresponding right whale habitat and quarter. Black borders around a box indicate that decade was a significant predictor of C. finmarchicus abundance. Positive numbers (blue boxes) correspond to areas where abundance was higher in the 2000s, and negative numbers (red boxes) correspond to areas where abundance was higher in the (A) 1990s or (B) 2010s. The color bar limits were set to range between 1 Â 10 5 and À1 Â 10 5 individuals per 10 m 2 , and dark red or dark blue squares may correspond to model predictions that exceed those values (see Tables S4, S5).
Finally, comparisons of late-stage C. hyperboreus abundance between the 2000s and the 2010s within Canadian waters using the AZMP data resulted in a total of 14 models, corresponding to the same quarters and sampling stations where models were run for C. finmarchicus. Nine of these models had a significant effect of decade and six significant models demonstrated a decrease in late-stage C. hyperboreus abundance in the 2010s relative to the 2000s (Fig.7B). Three of the significant models, all from the Rimouski station, demonstrated an increase in late-stage C. hyperboreus abundance in the 2010s relative to the 2000s (Fig. 7B). Among significant models, the predicted difference in late-stage C. hyperboreus abundance ranged from 4100 more organisms per m 2 in the 2010s (Rimouski, Quarter 3) to 3500 more organisms per m 2 in the 2000s (Shediac Valley, Quarter 3) (Fig. 7B). Model coefficients and model-predicted C. hyperboreus abundance values for the 2000s and 2010s for each station and quarter are provided in Supporting Information Table S7.

Relating right whale sightings to prey abundance
Comparisons between right whale habitat use and late-stage C. finmarchicus regional-scale abundance between the 1990s and 2000s provide insight into right whale behavior. Prey abundances were lower in most of the surveyed regions in the 1990s relative to the 2000s in the Gulf of Maine and Scotian Shelf regional high-use areas (Fig. 6A), coinciding with higher SPUE in the 1990s in the most reliable foraging habitats, such as Cape Cod Bay in the winter and Bay of Fundy in the summer, in contrast to broader use of the Gulf of Maine in the 2000s ( Fig. 2A, Fig. 3). Although late-stage C. finmarchicus abundance was significantly higher in the Bay of Fundy in the 2000s in quarters 2 and 3 (Fig. 6A), in the 1990s right whale SPUE was significantly higher in this habitat in June and July ( Fig. 2A). This may have occurred because relatively poor foraging conditions across the basin in the 1990s resulted in a concentration of right whales in Cape Cod Bay and the Bay of Fundy where chances of successful foraging were highest ( Figs. 2A, 3A).
Similar insights can be drawn from comparing right whale habitat use and prey abundance between the 2000s and 2010s. Higher late-stage C. finmarchicus abundances in Quarter 2 in Jeffreys Ledge, Cape Cod Bay, and the Great South Channel in the 2010s relative to the 2000s (Fig. 6B) corresponds with an increase in right whale SPUE in Cape Cod Bay and the adjacent Massachusetts Bay during April and May (Fig. 2B,  Fig. 4). The Quarter 3 decrease in late-stage C. finmarchicus in the Bay of Fundy and Gulf of Maine in the 2010s relative to the 2000s (Fig. 6B) may explain the decline in right whale SPUE in the Bay of Fundy in August, September, and November, as well as the new occupation of foraging grounds in the Gulf of St. Lawrence during summer and autumn (Figs. 2B, 5). Significantly higher prey abundances in Quarter 4 in the Gulf of Maine and Jeffreys Ledge in the 2000s (Fig. 6B) correspond directly with higher right whale SPUE in those regions compared to the 2010s (Fig. 2B).
The significant decline in late-stage C. finmarchicus (Fig. 7A) and C. hyperboreus (Fig. 7B) abundance during summer at the Shediac Valley AZMP station indicates that right whale use of the southern Gulf of St. Lawrence in the late 2010s is not driven by an increase in prey abundance in this more northerly habitat. Instead, as indicated by the significant summer late-stage C. finmarchicus decline at Halifax-2 (Fig. 7A), Bay of decades for each quarter at three AZMP stations. Box colors indicate differences in predicted abundance between the 2000s and 2010s for the median plankton tow start depth in the corresponding quarter and station. Black borders around a box indicate that decade was a significant predictor of copepod abundance in the associated model. Positive numbers (blue boxes) correspond to areas where abundance was higher in the 2000s, and negative numbers (red boxes) correspond to areas where abundance was higher in the 2010s.
Fundy, Gulf of Maine, Great South Channel, and Georges Bank (Fig. 6B), and the coherent decline at Prince-5 in the Bay of Fundy (Fig. 7A), right whales were most likely driven to higher use of the southern Gulf of St. Lawrence by inadequacies in prey availability in previously consistent high-use summer foraging habitats in the Gulf of Maine and Scotian Shelf. This evidence that right whales are being driven away from foraging habitats due to decreased prey availability, rather than being driven toward foraging habitats with increased prey availability, has important implications for our understanding of right whale habitat-use decision-making.

Discussion
Analysis of right whale sightings and Calanus spp. monitoring demonstrates significant changes in right whale spatial distribution and prey abundance over the previous three decades across the primary known range of the species. These results support the definition of a new pattern of right whale seasonal habitat use. While some patterns in seasonal habitat use remained consistent across all three decades, including the winter migration to the Southeast US calving ground and early spring foraging in Cape Cod Bay, there were marked differences in the seasonal timing and persistence of right whales in some foraging habitats across these three decades. In comparison to the 2000s, the decades of the 1990s and the 2010s were similar in that right whale SPUE declined in the Gulf of Maine and Scotian Shelf spring, summer, and autumn foraging habitats when abundance of C. finmarchicus was relatively low. The reduction in spatial extent of right whale SPUE, and in some years abandonment of important feeding grounds, is associated with longer duration of habitat use in Cape Cod Bay into the late spring, and increased use of Southern New England and the Gulf of St. Lawrence in the spring and summer in the 2010s. While rigorous survey effort in the latter half of the 2010s clearly demonstrates high right whale densities in the Gulf of St. Lawrence, the Mingan Island Cetacean Study (Fig. 5) and passive acoustic monitoring (Simard et al. 2019) time series provide evidence of increased usage of the Gulf of St. Lawrence since $ 2015.
Sightings from the Mingan Island Cetacean Study time series indicate that right whales have used the Gulf of St. Lawrence as a potential foraging refuge during previous periods of low prey abundance in the Gulf of Maine and on the Scotian Shelf. Right whale sightings in the Jacques Cartier Passage in 1994 and 1998 correspond to periods when prey abundance in the Gulf of Maine was anomalously low and right whales reduced their use of the Gulf of Maine and Scotian Shelf for summertime foraging (Kenney 2001;Patrician and Kenney 2010;Davies et al. 2015). Spatial climatologies of Calanus spp. biomass north of Anticosti Island suggest that this subregion may be a viable foraging habitat during summer and fall months . Since considerably fewer right whale sightings and individuals were observed in 1994 and 1998 in the Jacques Cartier Passage relative to the sightings in the 2010s, it is possible that the Gulf of St. Lawrence may not have been an adequate alternative foraging area in the 1990s. However, the Jacques Cartier Passage does not provide a complete representation of right whale habitat use in the Gulf of St. Lawrence. Current habitat-use patterns indicate that right whale detections have been primarily concentrated south of the Jacques Cartier Passage in the southwestern Gulf of St. Lawrence, especially near Shediac Valley (Simard et al. 2019;Crowe et al. 2021). No right whales were detected during aerial surveys that covered the entire Gulf of St Lawrence in August 1995(Kingsley and Reeves 1998Lawson and Gosselin 2009), although these surveys only included single passes over the Shediac Valley.
There is high variability in survey effort between and within right whale habitat polygons. By aggregating right whale sightings for a given month and polygon across each decade, survey effort was high enough to meaningfully assess decadal variation in seasonal habitat usage. However, decadalscale variation in right whale use of the Gulf of St. Lawrence is an exception to this, as effort in both the 1990s and 2000s was very low over a large spatial area. Survey effort increased starting in 2017 due to the discovery of 12 carcasses in the area, which indicated that habitat use was changing . The low survey effort in the previous decades is partially addressed by presenting sightings from Mingan Island Cetacean Study. Right whale habitat use may have also changed substantially between decades in other months and polygons than those presented here, but due to variable survey effort the data were not sufficient to robustly examine these patterns. Notably, variation between years but within a decade certainly occurs and is another important component of right whale habitat-use variation.
While some of the recent shifts in right whale detections have been described in the literature, including in Cape Cod Bay (Mayo et al. 2018;Charif et al. 2020), Southern New England (Quintana-Rizzo et al. 2021O'Brien et al. 2022), the Great South Channel (Record et al. 2019), Bay of Fundy Record et al. 2019), and the Gulf of St. Lawrence (Simard et al. 2019, Crowe et al. 2021, this is the first attempt to broadly characterize new seasonal habitat-use patterns across the core right whale range. This synoptic analysis provides the opportunity to connect emerging patterns of habitat use in one region with changes occurring in another region. In many cases, a decline in the use of one habitat can be associated with an increase in use in different habitat(s) during the same season. For example, the increased use of Cape Cod Bay and Massachusetts Bay in the 1990s and 2010s may be directly related to the reduction in migration of pregnant females and their companions to the Southeast US calving grounds (Fig. 2). Low prey abundance in both the 1990s and 2010s directly corresponded with a decline in right whale reproduction rates (Meyer-Gutbrod et al. 2015. With fewer births in these time periods, there is a larger subset of the population ready to get an early start on spring foraging. There also seems to be a shift from elevated right whale occurrence in the Bay of Fundy in the 1990s in summer and autumn to increased occurrence in the Great South Channel in the 2000s in summer and Roseway Basin in the 2000s in autumn. A similar transition appeared to occur with higher summertime sightings in most Gulf of Maine and Scotian Shelf regional high-use areas in the 2000s shifting to increased use of the Gulf of St. Lawrence in the 2010s; however, there was not enough right whale survey effort to confirm this pattern. These alternations between seasonal habitat preferences demonstrate that the models presented here are capturing meaningful patterns in habitat-use change, despite uneven survey effort. Right whale foraging habitat selection is driven by complex spatial and temporal patterns in environmental cues and prey abundance. For example, declines in C. finmarchicus abundance post-2010 were coherent in the Gulf of Maine except in portions of the western Gulf of Maine in Quarter 2 (Fig. 6B). This is consistent with recent positive anomalies of C. finmarchicus in the Wilkinson Basin area in Quarter 2 (Record et al. 2019;Ji et al. 2022), and is supported by the hypothesis that ecological processes influenced by the Maine Coastal Current are resilient to broader scale temperature increases (Ji et al. 2017;Runge et al. 2015). This relative increase in C. finmarchicus abundance appears to have supported increased occurrence and duration of right whales in Cape Cod Bay during spring in the 2010s (Fig. 2B, Record et al. 2019).
Recent summertime declines in Calanus spp. in the southern Gulf of St. Lawrence (Fig. 7, Sorochan et al. 2019) actually correspond with an increase in right whale occurrence (Fig. 2B, Fig. 4B, Crowe et al. 2021), likely because this habitat still serves as a foraging refuge when prey abundance in the Bay of Fundy and Roseway Basin habitats is low. Prey limitations driving reduced use of historic foraging habitats and the emergence of Gulf of St. Lawrence habitats are reflected in recent declines in right whale reproduction rates (Meyer-Gutbrod et al. 2021). Prey densities may not have been sufficient to support reproductive females in the southern Gulf of St. Lawrence initially (Gavrilchuk et al. 2021); however, the increase in calf births in 2021 may indicate that this region is becoming more suitable (Bishop et al. 2022;Pettis et al. 2022). Right whales may have needed time to acclimate to the new foraging environment in the Gulf of St. Lawrence (Pershing and Pendleton 2021), or increased Calanus spp. abundance in 2019 made this region more suitable for supporting reproduction (Blais et al. 2021).
Projections of novel oceanographic conditions and zooplankton distributions driven by anthropogenic climate change throughout the remainder of the 21 st century exacerbate challenges to the prediction of right whale distribution and demography (Ross et al. 2021). Climate projections indicate that C. finmarchicus distributions will continue to shift poleward in the coming decades (Reygondeau and Beaugrand 2011) leaving significantly reduced abundance in the Gulf of Maine (Grieve et al. 2017). However, due to the complex circulation patterns and bathymetry in this region, only climate models with sufficiently high spatial resolution to capture the position of the Gulf Stream and the deep entrance channels into the Gulf of Maine are able to reliably characterize oceanographic conditions (Saba et al. 2016;Seidov et al. 2021). Accurate projections of C. finmarchicus population levels under anthropogenic climate change are challenging; however, prey abundance alone is not sufficient to identify future right whale foraging areas. Projections must account for the physical and biological processes that aggregate zooplankton into sufficiently dense patches to meet right whale energetic requirements (Sorochan et al. 2021). To adapt to new environmental conditions, right whales may also target different prey species, including C. hyperboreus and C. glacialis in the Gulf of St. Lawrence. In this region, C. hyperboreus is likely an important prey item for right whales in spring and early summer (Lehoux et al. 2020). Analyses of regional variation in depth-integrated zooplankton abundance appear to provide a reasonable, but imperfect, proxy for suitable right whale foraging conditions (e.g., Pendleton et al. 2009;Record et al. 2019). Right whales rely on extremely high densities of prey to maximize their foraging efficiency (Kenney et al. 1986) at depths that are energetically feasible for foraging dives (e.g., Gavrilchuck et al. 2020). These prey aggregations are formed by processes that influence prey supply (e.g., vital rates and advection) and local-scale biophysical prey-concentrating mechanisms (Sorochan et al. 2021). Interpretation of suitable foraging habitat from depth-integrated abundance data is limited because a high abundance value could be generated by a diffuse vertical distribution. Time series of depth-integrated and spatially and temporally aggregated abundance are effective at capturing variation caused by various supply processes at the regional scale (e.g., Ji et al. 2022), but not prey-concentrating mechanisms at the local scale. Since right whales are both highly mobile and responsive to changes in foraging conditions, temporal resolution of prey availability on the scale of 3 months (quarters) does not capture fine-scale temporal variation relevant to some right whale habitat-use preferences.
While gaps remain in coverage of right whale movements throughout the year, this analysis shows that many major right whale habitats are regularly monitored, and shifts in spatial and seasonal distribution can be resolved. With increased survey effort in the Gulf of Maine in the 2000s and now the Gulf of St. Lawrence in the 2010s, scientists have begun to unravel the mystery of wintertime distribution by identifying habitats used by a subset of the right whale population. Since 2000, between 57% and 95% of right whale individuals were sighted annually, with the lowest sighting rates occurring between 2012 and 2016 (Pettis et al. 2022). However, the majority of right whale deaths are unobserved (Pace et al. 2021). There are three possible scenarios to explain the unobserved portion of the population in a given year: (1) the major seasonal habitats have been identified, but effort is not high enough to observe all of the individuals within those habitats; (2) there exists critical, high-density right whale habitats that have not yet been identified; (3) individuals that have not been observed in the major habitats are spread out diffusely throughout the Northwest Atlantic, exhibiting more isolated behaviors. Scenarios (2) or (3) might explain the relatively low number of right whales seen in 2012-2016 when patterns in right whale habitat use appeared to shift (Pettis et al. 2022). Uncertainty over the distribution of the unobserved population impairs the implementation of effective protective policies to reduce vessel strikes and fishing gear entanglement. The potential existence of high densities of right whales in unregulated habitats, similar to the Gulf of St. Lawrence in 2015-2017Davies and Brillant 2019), may be the tipping point between the recovery and the extinction of this species. Further efforts to resolve changes in right whale seasonal habitat-use patterns, including expanded monitoring and predictive modeling, are essential to protect the endangered North Atlantic right whale while accommodating those industries that put them at risk.

Data availability statement
Right whale population data are available from the North Atlantic Right Whale Consortium. The Calanus finmarchicus abundance data collected from the EcoMon survey program are publicly accessible at the NOAA National Centers for Environmental Information site (https://www.ncei.noaa.gov/ archive/accession/0187513). The AZMP Calanus abundance data are available from the Government of Canada Open Data website (Prince-5 and Halifax-2) and from Research Document 2021/060 (dfo-mpo.gc.ca) (Shediac Valley and Rimouski).