Foraging habitat of North Atlantic right whales has declined in the Gulf of St. Lawrence, Canada, and may be insufficient for successful reproduction

Climate-induced changes in calanoid copepod (Calanus spp.) availability in traditional feeding areas might explain why a large proportion of the North Atlantic right whale Eubalaena glacialis population has fed in the Gulf of St. Lawrence (Canada) in recent years. However, little is known about the distribution of copepods in the gulf, and whether their abundance is sufficient to energetically sustain right whales. We used a mechanistic modelling approach to predict areas within the gulf that have foraging potential for adult female right whales, based on the annual energetic needs of resting, pregnant and lactating females, and their theoretical prey density requirements. We identified suitable foraging areas for right whales by coupling a foraging bioenergetics model with a 12 yr data set (2006−2017) describing the abundance and 3-dimensional distribution of late-stage Calanus spp. in the gulf. Prey densities in the southern gulf (from Shediac Valley to the Magdalen Islands) supported all 3 reproductive states in most (≥6) years. However, foraging habitat became progressively sparse in the southern gulf over time, with noticeably less suitable habitat available after 2014. Few other potentially suitable foraging areas were identified elsewhere in the gulf. Overall, the availability of foraging habitat in the gulf varied considerably between years, and was higher for resting females than for pregnant and lactating females. Our findings are consistent with the recent low calving rates, and indicate that prey biomass in the Gulf of St. Lawrence may be insufficient in most years to support successful reproduction of North Atlantic right whales.


INTRODUCTION
A central objective of applied biological conservation is to identify and control potential threats in important geographic areas used by vulnerable species. Important areas for vulnerable species (or populations) can be identified using correlative or mechanistic modelling approaches (Kearney et al. 2010). Correlative modelling approaches attempt to link species occurrence with simultaneously collected environmental variables and predict species distribution using associated spatial correlates (Guisan & Zimmermann 2000). In contrast, mechanistic or bottom-up modelling approaches evaluate the biophysiological traits of a species to determine which habitat conditions likely favour species occurrence or fitness (Kearney & Porter 2009). Mechanistic habitat modelling seeks to understand the latent processes responsible for animal use of a given area, and is often rooted in first principles of physiology and biome-chanics, which govern all organisms (Peterson et al. 2015). The fundamental unit of measure in such models is energy (expressed in J), known as the universal currency in ecology for its direct influence on survival and reproduction (Stephens & Krebs 1986, MacNab 2002. Bioenergetic habitat suitability models are mechanistic in nature, incorporating physiological traits and energy requirements of organisms to better understand patterns of habitat use (van der Vaart et al. 2016). Such models are a valuable tool in wildlife research and management, and have been used to identify suitable habitat for a variety of species globally (e.g. Tomlinson et al. 2014, Silber et al. 2017. Understanding the energy requirements of a species can further help predict future distributions, and evaluate the energetic costs associated with anthropogenic disturbance that reduce foraging success and ultimately impede population growth or recovery (Costa 2012).
From a reductionist perspective, the likelihood of an animal feeding at a given location reflects the trade-off between energy expended and energy gained, where animals are assumed to favour food items or foraging sites that provide a gain in net energy over the long term, needed to sustain life history functions and vital rates (Hall et al. 1992, Olsson & Bolin 2014. In general, energy intake depends on body size, physiological requirements, nutritional status and the environmental factors influencing the spatial distribution, availability, quantity and quality of food sources (Chipps & Wahl 2008). In turn, energy expenditure depends on physiological requirements such as basal maintenance, growth, gestation, lactation, thermoregulation, digestion and locomotion (Sibly & Calow 1986). Consequently, a particular habitat may not be equally suitable for all demographic groups of a population. Bioenergetic models can therefore be generated with special focus on susceptible demographic groups, such as reproductively mature females that have higher energetic burdens and play a critical role in supporting newborn survival (Fortune et al. 2013, New et al. 2013, Villegas-Amtmann et al. 2015. The North Atlantic right whale Eubalaena glacialis (hereafter right whale) is a Critically Endangered balaenid cetacean (Cooke 2020), distributed primarily in coastal waters of the western North Atlantic, from the southeastern USA to eastern Canada (COSEWIC 2013). The population numbered between 444 and 471 individuals in 2015 (Pace et al. 2017), but has since declined due to a combination of low calving rates and high mortality rates (Pettis et al. 2018. Mortality caused by fishing gear entanglement and vessel strikes has been the primary driver of population decline (Corkeron et al. 2018, Sharp et al. 2019. A portion of the right whale population, comprised mostly of reproductively mature females, calves and some juveniles, migrates annually between winter calving areas off Georgia and Florida (USA) and summer feeding areas in the northern USA and Canada (Gowan et al. 2019). Since the early 1980s, around two-thirds of the population occupied the lower Bay of Fundy and the Scotian Shelf during summer and fall. Small numbers of animals also fed in the Gulf of St. Lawrence (COSEWIC 2013; Fig. 1). However, seasonal occurrence patterns have noticeably shifted during the last decade, with long-term acoustic detections indicating that right whales occur less frequently in the Bay of Fundy and the greater Gulf of Maine, and more frequently in southern New England and the mid-Atlantic region year-round (Davis et al. 2017). This change in distribution is also supported by long-term sightings data that report significantly fewer right whales on traditional, highuse feeding grounds such as the Bay of Fundy over the past decade (Davies et al. 2019). Passive acoustic monitoring has also detected an increase in summertime use of the Gulf of St. Lawrence by right whales since at least 2015 (Simard et al. 2019).
The shift in distribution of right whales away from their traditional feeding grounds may reflect a climate-induced change in the distribution and availability of prey (Meyer-Gutbrod et al. 2015, Record et al. 2019. Warming ocean temperatures may continue to move right whales further north given the northward range shifts observed in some forage species, and the impact that warmer waters throughout the Northeast US Shelf are predicted to have on the distribution and abundance of zooplankton (Reygondeau & Beaugrand 2011, Grieve et al. 2017. Such changes in prey availability may also result in some right whales forgoing migration to wintering grounds off the southeastern USA in years of reduced prey availability (Krzystan et al. 2018, Gowan et al. 2019. Larger-sized calanoid copepods such as Calanus finmarchicus, C. glacialis and C. hyperboreus dominate the total mesozooplankton biomass in the western North Atlantic (Pepin et al. 2015). Right whales have been observed to primarily forage on calanoid copepods such as late-stage C. finmarchicus, Pseudo calanus spp. and Centropages spp. (reviewed by Baumgartner & Mate 2003) and occasionally on euphausiids (see Murison & Gaskin 1989). However, right whale baleen is most efficient at filtering the larger, late-stage (CIV to adult) copepods (Mayo et al. 2001). Given the strong correlation between C. fin marchicus abundance in the Gulf of Maine and right whale body condition and calving rates (reviewed by Meyer-Gutbrod & Greene 2018), it is reasonable to suspect that shifts in right whale distribution during the feeding season reflect changes in the availability and/or distribution of C. finmarchicus. Predicting which new geographical areas may attract right whales based on their energetic requirements and the distribution of their preferred prey is one means to inform future monitoring efforts and conservation of this Critically Endangered species.
A preceding study evaluated areas of suitable foraging habitat for right whales in the Gulf of St. Lawrence, Scotian Shelf and Bay of Fundy by applying basic energy requirements of an average right whale to a 3-dimensional (3D) spatial climatology of multi-year average Calanus spp. densities . Here, we built on the work of Plourde et al. (2019) by assessing interannual variation in the 3D Calanus spp. prey field and increasing the resolution of the bioenergetics model. We also predicted the energy requirements of reproductively mature adult females in order to assess the suitability of the Gulf of St. Lawrence to meet their foraging needs. This study supports future survey planning and identification of areas that may benefit from adaptive management approaches to protect right whales from human-caused mortality in eastern Canadian waters.

Data collection
Zooplankton sampling surveys conducted in the Estuary and Gulf of St. Lawrence since 1979 have yielded a prey database with broad spatial and temporal coverage . The term Gulf of St. Lawrence or gulf will refer to both the Estuary and Gulf regions combined from here on. For our study, we selected sampling years with the most consistent spatial coverage (2006 to 2017) during the summer period (June to September), a period that coincides with the peak abundance of right whales in this area (DFO 2019a). Early summer surveys (June to early July) were primarily carried out in the south- ern gulf following a systematic design, while the late summer surveys (late July to September) covered mainly the northern gulf and followed a random sampling de sign (Fig. 2). Sampling surveys were conducted in the absence of foraging right whales.
The methodology for depth-integrated zooplankton sampling and abundance assessment (ind. m −2 ) for the 3 dominant Calanus copepod species (C. finmarchicus, C. glacialis and C. hyperboreus) was summarized by Plourde et al. (2019) and Sorochan et al. (2019). Briefly, samples were collected with oblique or vertical tows respectively performed with Bongo nets (opening 0.61 m; mesh size 333 μm) or ring nets (opening 0.75 m; mesh size 200 μm). Samples were analysed using a semi-automated image analysis procedure for species and life-stage identification and enumeration, resulting in 2 copepod classes that encompassed multiple copepodite sta ges: C. finmarchicus and C. glacialis CIV−VI, and C. hyperboreus CIV−VI. These copepod classes overwhelmingly dominated the size fraction >1 mm, and were considered a good proxy of Calanus spp. biomass. The semi-automated estimates of in situ Cala nus spp. abundance were then corrected based on the relationship between taxonomist-measured abundance and semi-automated analysis in a subset of samples

Prey field
Following the steps described in Table 1, we generated a continuous prey field over the study area based on depth-integrated Calanus spp. dry weight (DW) biomass (g DW m −2 ) measured at discrete sampling stations in the Gulf of St. Lawrence over the 12 yr study period (2006−2017). We first used Bayesian geostatistical inference to predict the 2-dimensional (2D), depth-integrated prey field for both early and late summer of each year (e.g. Cosandey-Godin et al. 2015). We included potential static (bathymetry and slope) and dynamic (sea surface temperature, SST) environmental correlates known to affect zooplankton abundance to explore their utility in informing Calanus spp. biomass predictions (Albouy-Boyer et al. 2016, Plourde et al. 2016. Bathymetry (or depth, m) and slope (°) data for each sampling station was obtained from the satellite-derived, high-resolution (1 km 2 grid) MARSPEC database (www.marspec. org). Daily SST data were extracted via the ERDDAP data server from the Multi-scale Ultra-high Resolution (1 km 2 ) SST Global database (NASA JPL) (originally downloaded from https:// coastwatch. pfeg. noaa. gov/ erddap/griddap/jplMURSST.html). We used continuous domain stochastic partial differential equations to solve the covariance spatial structure of Calanus spp. biomass measurements (Lindgren et al. 2011) using the R package R-INLA (www. r-inla.org). Given that our response variable (Calanus spp. biomass) was positive-continuous with a small percentage (1%) of zeros, to avoid removing the zero data, we added a negligibly small value (0.001) to the zeros and evaluated the goodness of fit of 3 can didate distributions (gamma, exponential and lognormal).
The relationship between Calanus spp. biomass measurements and environmental correlates was modelled separately for each year, and for early summer (June to early July) and late summer (late July to September) (Eq. 1). The best-performing model (based on goodness of fit and predictive performance) was then used to predict Calanus spp. biomass at unsampled locations in our study area. To begin, we let y(s) denote the realisation (i.e. one set of possible outcomes) of a stochastic spatial process Y(s), which corresponds to our Calanus spp. biomass measurements (y 1 ,…, y N ) sampled at locations s 1 ,…, s N , such that: where location s is a spatial index of latitude− longitude values, z(s i ) is a vector of covariate values for location s i , and β is a vector of regression coefficients. The w(s i ) are the spatially correlated random effects, which capture any spatial pseudoreplication in the data, and are represented by a Gaussian Markov random field with mean 0 and covariance matrix Σ where the Matérn correlation function is used to parameterise the covariance matrix. The ε(s i ) are the independently distributed, spatially uncorrelated regression residuals where ε(s i )~ N(0,σ 2 ε ) (Lind gren et al. , Cameletti et al. 2013). All covariates were in cluded as smoothed (non-linear) effects given the non-linear relationships observed with the response variable during data exploration. All environmental correlates were standardized to avoid numerical esti- mation issues and to facilitate interpretation of regression parameters. Under the Bayesian paradigm, model parameters are considered random variables and defined using prior probability distributions. For the smoothed effect of the covariates, we assigned a first-order random walk model (Krainski et al. 2018) with penalized complexity (PC) prior probability distributions recommended by Simpson et al. (2017), where the probability of the covariates prior standard deviation (σ c ) being >1 is 0.05. For the spatial random effect, we set PC priors for the spatial correlation range r and σ r so that P(r < 10) = 0.05 and the P (σ r > 3) = 0.05 (Fuglstad et al. 2019).
We compared various candidate models (all additive covariate combinations) fitted with and without the spatial random effect using the deviance information criterion (DIC, Spiegelhalter et al. 2002) and the Watanabe-Akaike information criterion (WAIC, Watanabe 2010), where lower DIC or WAIC values indicate higher goodness of fit. Two additional crossvalidation methods were used to assess predictive performance of candidate models. We first randomly selected 80% of the stations for model estimation and 20% for validation to evaluate how well the model predicted Calanus spp. biomass at validation stations based on the correlation between observed and predicted Calanus spp. biomass. We then used the builtin R-INLA leave-one-out cross-validation diagnostic (the conditional predictive ordinate, CPO), which calculates a prediction error score for each observation. To summarize CPO values for all observations into 1 metric, we computed the sum of all log CPO values, for which larger values indicate better model predictive power (Gneiting & Raftery 2007).
We then used the final model to predict depthintegrated Calanus spp. biomass on a 10 × 10 km regular grid covering the study area in early summer (15 June) and late summer (15 August) for each study year. We extended our predictions to a maximum distance of 30 km from sampling sites to reduce uncertainty, and obtained covariate values at prediction locations the same way as for the observed Calanus spp. data.
Preceding studies assessed the seasonal vertical distribution of Calanus copepods in the western North Atlantic using depth-stratified abundance and biomass data, and a generalized additive modelling approach (Krumhansl et al. 2018). One of their resulting models predicts the daily averaged relative proportion (%) of depth-integrated total Calanus spp. biomass in 10 m depth intervals from surface to sea floor, capturing the seasonal changes in Calanus spp. vertical distribution associated with the transition between the active (0−100 m; late spring and early summer) and diapausing (>100 m; late summer) phases of their life cycle ). This model adequately described the concentrating effect on Calanus spp. biomass density (g m −3 ) in areas where the sea floor is shallower (<100 m) than the optimal overwintering depth (>150 m) selected by Calanus spp. in deep areas . We therefore applied this Calanus spp. biomass relative (%) vertical distribution model to our 2D 10 km 2 grid of predicted depthintegrated Calanus spp. biomass (g m −2 ) to produce a 3D representation of the prey density (g m −3 ) field.

Net energy field
From the 3D prey field, we extracted 2 parameters for the foraging bioenergetics model described below: Calanus spp. biomass density (g m −3 ) predicted in 10 m depth bins, and the depth of each layer or transit distance from the water surface to access prey. We then applied the foraging bioenergetics model to the 3D prey field by evaluating the balance between energy gain and expenditure for a given prey biomass density in a particular 10 m depth bin. This produced a 3D net energy field, where each 10 km 2 xyz location (latitude− longitude−depth) had a net energy index (E net ) represented by the following simple theoretical energy balance (Eq. 2): (2) where E net is the proportion of energy gained (E in , Eq. 3) or expended (E out , Eq. 4), relative to energy expended (E out ). An E net value of 0 signifies that energy gained equals energy expended. We considered E net = 0 to be the minimum theoretical foraging threshold. Any xyz locations with E net values > 0 were defined as 'suitable' for meeting the daily energy requirements of right whales based on the foraging assumptions made below. Values for model parameters are summarized in Table 2.

Energy gain
Right whales feed by ram filtration, which involves passively filtering large volumes of prey-filled water by slowly swimming forward with their mouth open (Mayo & Marx 1990 which depends on mouth opening area (A m , m 2 ), swim speed during feeding (U b , m s −1 ), time spent ingesting prey (T b , s d −1 ), prey energy content (E p , MJ g −1 ), prey density (D p , g m −3 ) and dietary assimilation efficiency (ε A ), or the ratio of energy gained to energy ingested after digestion (Table 2, see Text S1 in the Supplement at www.int-res.com/articles/ suppl/n044 p113_ supp.pdf for justification of parameter values).

Energy expenditure
We evaluated the annual 'cost of living' for an adult female North Atlantic right whale in either a resting (i.e. reproductively mature but not pregnant or lactating), pregnant or lactating state that forages during the summer, breeds or calves in the winter and migrates during the fall and spring. Using available literature on the general activity budget of North Atlantic right whales (Table 3, and references therein), we created a time−activity budget in which the proportion of time a right whale spends in 3 principal activity states (foraging, resting and travelling/ socializing) varied with season. Foraging was set as the main activity during the summer and was considered negligible during migration and while on the wintering grounds. Travelling was set as the main activity during the fall and spring migrations, and travelling and/or socializing as the main activities during winter (Table 3). Given the variability in swimming parameters and body kinematics during social activity, as well as the uncertainty in how to quantify associated energetic expenditure, we group -  (4) where i is the season (n = 4; winter, spring, summer, fall), a is the activity type (n = 3; foraging, travelling/ socializing, resting), T ai is the number of daily hours spent in activity a during season i, M a is the activityspecific energy cost, E repro is the energy required for gestation or lactation per year, and T f is the number of days spent foraging per year, inferred from observed right whale migration time and residency on wintering grounds (Table 3).
We used resting metabolic rate (RMR) as a proxy for basal metabolic rate (BMR) because BMR is difficult to measure under Kleiber standards for most marine mammals (e.g. ensuring individuals are motionless, thermoneutral, post-absorptive). We thus used the following allometric RMR relationship specific to marine mammals (Williams & Maresh 2015) that is comparable to the widely-used BMR equation of Kleiber (1975) where body mass is in kg, and RMR is in kJ d −1 (converted to MJ). An average body mass of 35 000 kg was used for a 14 m long right whale (~18 yr old; Moore et al. 2005). We held RMR constant because the magnitude of intraspecific variation in RMR for a given body mass is unknown for large whales. However, we acknowledge that more information on RMR in right whales would help refine future bioenergetic models. The cost of travelling and foraging was estimated using published morphometric and kinematic parameters for North Atlantic right whales (Table 2). Four main forces act on pelagic swimming species: weight, buoyancy, propulsion (or thrust) and drag, where weight counteracts buoyancy and propulsion counteracts drag (reviewed by Fish 2000). Buoyancy force fluctuates with individual body condition, and can cycle through periods of higher buoyancy during blubber accumulation and lower buoyancy during blubber breakdown (Nousek-McGregor et al. 2014). To quantify buoyancy force, data are needed on animal blubber volume, tissue density, and volume of gas-filled cavities (Miller et al. 2004). Since this infor-   Vermeulen et al. (2012): inferred from southern right whale Eubalaena australis activity budget on breeding ground. Travel (41%) and social (42%) activity grouped together and resting (11%) and 'other' (6%) activity grouped together.
Nousek McGregor (2010) found that tagged right whales on breeding grounds off Florida spend 64.7% of their time (174/269 dives) presumably travelling (deeper and longer dives) and 35.3% (95/269) of their time either socializing or resting (shallower and shorter dives) f Assumption Table 3. Inferred daily and seasonal time−activity budget for adult female North Atlantic right whales (NARWs). Note that annual phase durations are identical for resting and pregnant females. Travelling and socializing behaviour are grouped together under 'Travelling' for energy expenditure calculation mation was unavailable, we assumed that the energy expended by a positively-buoyant right whale to overcome buoyancy during dive descent by fluking is balanced by energy savings while gliding during ascent (Nousek-McGregor et al. 2014).
Propulsive power (P p in J s −1 , Eq. 6) depends on the hydrodynamic drag force (F drag in N), swim speed (U, m s −1 ) and energy transfer efficiency which includes both muscular (or metabolic) efficiency (η m ) and propulsive (or propeller) efficiency (η p ). It differs with activity (e.g. travelling vs. foraging) and dive phase (descent, bottom, ascent and surface recovery), and can be multiplied by the estimated time (s) a whale spends in each activity per day (Table 3) to determine daily activity costs (J d −1 ). We calculated P p separately for travelling and for each phase of a foraging dive using Eq. (6): (6) where the magnitude of the drag force (F drag , Eq. 7) depends on swim speed, body form, characteristics of the surrounding medium, proximity to boundaries such as the air−water interface, and the relative contribution of inertial, viscous and gravitational forces (Fish 1998). F drag on a moving right whale can be calculated as: (7) where ρ is the density of the surrounding medium (seawater), C D is the dimensionless drag coefficient (see Eq. 8), and A is the total wetted surface area (m 2 ), which can be approximated using an allometric relationship to body mass (Table 2). We increased the surface area A by 5% for pregnant females based on aerial photogrammetric measurements and blubber thickness data (Nousek McGregor 2010). U is the swim speed (m s −1 ), g is the added drag of appendages (flukes and flippers), λ is the active-to-passive drag ratio, and γ is the surface wave drag augmentation factor. For the latter 2 factors, body oscillation during active locomotion alters the hydrodynamics and drag regime around a moving body, and is quantified by λ. Whether active swimming increases or decreases overall drag relative to passive gliding in right whales is uncertain; we therefore assumed a conservative λ = 1, comparable to previous bioenergetic studies on marine mammals (Aoki et al. 2011, P. Miller et al. 2012, Trassinelli 2016. Surface wave drag (γ) affects animals swimming at or near the water surface and gradually decreases with depth, becoming negligible (γ = 1) at submergence depths greater than 3 times the maximum body diameter (Hertel 1969). Given the uncertainty regarding the influence of surface wave drag on right whale energy expenditure, we set γ = 1 for our model.
We calculated the drag coefficient (C D ) by multiplying the skin friction coefficient (first expression in Eq. 8) by the body dimensions, also known as the 'form factor' (second expression in Eq. 8; Phillips et al. 2017): (8) where L is the body length (m), d is the maximum body diameter (m), and Re is the Reynold's number (Reynolds 1883;Eq. 9), which describes the flow regime around a body of length L moving at speed U (m s −1 ) through a medium with kinematic viscosity ν ( Table 2): (9) At high Reynold's numbers (> 5 × 10 6 ), which is the case for right whales given their body size and typical swim speed, fluid flow around the body transitions from laminar to turbulent, and increases the value of the skin friction coefficient (Kline et al. 1967). Under a turbulent flow regime, the skin friction component of the drag coefficient (C D ) can be calculated using the constant α = 0.072 and the exponent β = −0.2 (as used by Phillips et al. 2017).
The drag force acting on a right whale during foraging is greater than during travelling, since openmouth feeding exposes a larger surface area to oncoming water flow, adding resistance to forward mo tion (Sanderson & Wassersug 1990. For a given speed, foraging increases the drag on a right whale by an estimated 2−3 times compared to travelling (Nousek McGregor 2010. F drag (Eq. 7) was thus multiplied by 2 and 3 to obtain low and high estimates of energy expended during foraging.
Most deep-diving marine mammals use an intermittent swimming gait to conserve oxygen reserves, where active stroking is followed by periods of passive gliding. Certain species also increase time spent gliding during descent on deeper dives (Williams et al. 2000). On a foraging dive, right whales passively glide for an average 36% of descent time (range 10−75%), 30% of ascent time (20−36%) and 9% of bottom time (6−12%; Nowacek et al. 2001, Nousek McGregor 2010, Nousek-McGregor et al. 2014. The proportion of time right whales spend gliding during surface recovery and during travelling is unknown, so we assumed a similar glide time as during horizontal swimming during the bottom phase (9%). We used average glide times and kept the proportion of time spent gliding in each phase constant with depth, as it is unknown how glide times vary with dive depth in right whales. The metabolic cost incurred during gliding likely falls between RMR and active propulsion. However, in the absence of data to in form the cost of this locomotory behaviour, we as sumed energy expended during a passive glide is equivalent to RMR, recognizing that this may be an underestimate. Therefore, the total energetic cost (J, converted to MJ) of travelling per day was calculated as: and the total cost (J, converted to MJ) of foraging per day at depth i as: where j is 1 of 4 dive phases, P p is the propulsive power (J s −1 ) during dive phase j, t travel and t j are the times (s) spent travelling per day or in dive phase j per dive, t glide is the proportion of time spent gliding. n dives is the theoretical maximum number of dives d −1 at a given depth if foraging ~15−17 h d −1 (Table 3), which would translate to daily ingestion (bottom) times of ~5−10 h d −1 , depending on reproductive state, daily foraging time and foraging depth. This is in accordance with an average daily ingestion time of 8.4 h d −1 reported from tagged North Atlantic right whales (Goodyear 1996). Time (s) spent in ascent and descent phase was calculated as distance travelled (m) divided by swim speed (m s −1 ) for each phase, where distance travelled is equal to dive depth (m) corrected for average body pitch (θ) (Eq. 12). During foraging dives, adult female right whales pivot their body to around 69−78° relative to the water surface on descent, and 56−68° relative to the sea floor on ascent, depending in part on body condition (Nousek-McGregor et al. 2014). We used average angles of 74° for descent and 62° for ascent, and calculated distance travelled during descent or ascent phase as: Bottom time for dive depths of 160−500 m was assumed constant and equal to the maximum bottom time predicted by the previous equation (t = 636 s). Time spent in the surface phase (t s ) was expressed as a percentage of the total time spent at depth, and was assigned an average value of 21.1% for resting females and 34.2% for pregnant or lactating females (Baumgartner & Mate 2003).
The gestation period for right whales is about 12 mo (Knowlton et al. 1994). During this time, females allocate energy to foetal, uterine, placental and ma mmary tissue growth. Following parturition, mothers need energy to produce milk and provision their young until weaning. The maternal energy required for gestation is allometrically correlated with newborn body mass (Eq. 14), and this relationship has been used in several studies on cetacean energetics (e.g. Lockyer 1981, Fortune et al. 2013, New et al. 2013, Villegas-Amtmann et al. 2015: where H g is the heat increment of gestation (in kCal, converted to MJ using 1 kCal = 0.004184 MJ), and mass nb is newborn mass (kg).
The energetic cost of lactation is a combination of both mother and calf metabolism (Lockyer 1981). The nursing period for large mysticetes is around 6−7 mo (Oftedal 1997), and North Atlantic right whale mothers and calves remain together for 8−17 mo (Hamilton et al. 1995, Hamilton & Cooper 2010. Calves may begin feeding on zooplankton as early as 8 mo old (Baumgartner & Mate 2003). We estimated the cost of lactation by adding the average daily energy requirement of the calf (mean ± SD = 1767 ± 261 MJ; Fortune et al. 2013) to the daily energy requirement of a lactating female, accounting for a milk transfer efficiency of 90% (Lockyer 1981). We assumed mothers provided 110% of the daily energy requirements of the calf for the first 6 mo after birth, followed by a linear decline in energy contribution (representing the gradual weaning period) up until 12 mo, at which point the calf was presumed to feed independently (New et al. 2013).

Foraging habitat suitability
We converted the 3D net energy field into a foraging habitat suitability grid for each year and reproductive state, as well as for all years combined. We θ descent or ascent Depth Distance = first summed the number of 10 m depth bins with E net values > 0 (where 0 is defined as energetic equilibrium) for each grid cell, then divided by the total number of depth bins per grid cell. This provided a relative and weighted measure of foraging value per cell, with a value of 1 signifying that 100% of the water column was predicted to have suitable prey densities. We then built composite maps by overlaying all 12 annual habitat suitability maps together (2006−2017), and counting the number of years each 10 km 2 grid cell had a positive foraging value (comparable to Nelson et al. 2009). This provided information on which areas in the Gulf of St. Lawrence showed temporally persistent suitable habitat (defined as any xyz 10 km 2 location within the gulf where prey density exceeded the theoretical daily energy requirements for adult female North Atlantic right whales). We considered 2 scenarios which differed in terms of energy gain and expenditure, reflecting suboptimal (low energy gain for high energy expenditure) and optimal (high energy gain for low energy expenditure) habitat suitability in the Gulf of St. Lawrence (Tables 2 & 3; Table S1). Trends over the study period in the amount of suitable foraging habitat (i.e. the number of 10 km 2 grid cells with at least one 10 m depth bin showing an E net > 0, corrected for sampling effort) were explored using Pearson's correlations. All analyses were performed in R version 3.6.1 (R Core Team 2020).

RESULTS
In total, 1543 stations were sampled for Calanus spp. abundance from 2006 to 2017 in the Estuary and Gulf of St. Lawrence. Sampling effort during early summer surveys of the southern gulf was relatively uniform across years. Effort was also relatively uniform during the late summer surveys from 2006 to 2014 primarily in the northern gulf, but was low in 2015 and 2017, and nonexistent in 2016 (Fig. 2, Table S2).
When predicting the 2D depth-integrated seasonal prey fields, the gamma distribution outperformed the exponential and lognormal distributions for the early-season Calanus spp. biomass, and all 3 distributions were comparable in terms of model fit and predictive performance for the late-season biomass data. The gamma distribution was chosen for prediction with the best-fitting model, which included additive smoothed effects of SST, slope, bathymetry and the spatial random effect.
Using the estimation and validation data subsets, observed depth-integrated Calanus spp. biomass values were correlated to predicted values with a mean ± SD correlation coefficient of 0.61 ± 0.11 in the early season and 0.69 ± 0.10 in the late season. SST and slope did not significantly affect Calanus spp. biomass during the early or late season (95% confidence intervals overlapped 0 in all years). Unsurprisingly, bathymetry had little effect on Calanus spp. biomass values in the early season when 90% of the samples were collected in the southern Gulf of St. Lawrence at depths <100 m (not shown). However, a positive effect was detected in the late season when Calanus spp. were sampled across a wider range of depths (40−450 m) (Fig. S1). We found a strong interannual variability in spatial correlation of Calanus spp. biomass values, with mean ranges of 71−434 km (median: 219, mean: 202, SD: 106) in the early season, and 53−1642 km (median: 133, mean: 376, SD: 506 km) in the late season depending on the year. The degree of autocorrelation and uncertainty in predictions, expressed as the mean and SD posterior values for the spatial random effect, varied spatially and interannually (Figs. S2 & S3).
Annual energy expenditure (MJ d −1 ) estimated for an adult female North Atlantic right whale in either a resting, pregnant or lactating reproductive state is summarized in Table 4. Assuming right whales spend 15−17 h d −1 foraging during the summer, Fig. S4 shows the predicted daily ingestion time (h) when foraging at depths from 0 to 500 m. It also shows the minimum predicted prey density requirement as a function of depth for each reproductive state. Prey biomass densities required to meet daily energy output ranged from 0.62 to 2.67 (median 1.34) g m −3 for resting females, from 0.79 to 3.72 (median 1.80) g m −3 for pregnant whales and from 1.82 to 8.48 (median 4.11) g m −3 for lactating whales. This translates to threshold abundance densities of ~4470− 6700 ind. m −3 for resting, ~6000−9000 ind. m −3 for pregnant and 13 700−20 550 ind. m −3 for lactating females if they fed only on C. finmarchicus/ C. gla cialis stage CV. Alternatively, required threshold abundances would be ~610−790 ind. m −3 for resting, ~820− 1060 ind. m −3 for pregnant, and ~1870− 2420 for lactating females feeding only on C. hyperboreus stage CI−V (using our median prey biomass density thresholds and the range of individual copepod dry weights summarized by Lehoux et al. 2020).
The depths associated with maximum E net values (≥ 98 th percentile) to support all 3 reproductive states of right whales were 50 m (median) in the early summer, and 220 m in late summer. These coincided with maximum Calanus spp. densities at 50 m in early summer in the shallow southern Gulf of St. Lawrence, and 230 m in late summer (when sampling effort was mainly in the deeper northern Gulf of St. Lawrence). The total percentage of E net values exceeding 0 within the 3D Calanus prey field was low, ranging from 1.0 to 6.5% for resting females, from 0.5 to 4.7% for pregnant females and from 0.01 to 0.9% for lactating females.
The location of suitable foraging habitat varied considerably from 2006 to 2017 (Fig. S5). As ex pected, more habitat was available to right whales under the optimal foraging scenario compared to the suboptimal scenario, and the amount of suitable habitat decreased as females went from resting, to pregnant, to lactating and their energy expenditure increased.
Suitable foraging habitat which met the energetic requirements of right whales in any reproductive state and across several years occurred in the southern Gulf of St. Lawrence. However, the spatial configuration and quantity of this habitat within the southern gulf fluctuated from year to year. Under an optimal foraging scenario, densities required to meet the needs of females with the lowest energy demands (i.e. resting females) occurred in ≥ 50% of the water column within some regions of the southern gulf over multiple years. These areas included Chaleur Bay, Shediac Valley and surrounding waters, the southern slope of the Laurentian Channel (to the north of the southern gulf), waters off the Gaspé Peninsula and around the Magdalen Islands, and the middle of the Magdalen Shallows (Fig. S5b).
Suitable foraging habitat became progressively more sparse from 2006 to 2017, with little to no suitable habitat available after 2014 for all reproductive states unless foraging conditions were optimal ( Fig. 3; Fig. S5). This pattern was particularly striking in the southern Gulf of St. Lawrence, with a significant decline in habitat suitability under both foraging scenarios (Pearson's R, all > 0.63, all p < 0.05), when suitable habitat was available (i.e. none available for the 2006−2017 period for lactating females foraging under suboptimal conditions). In the northern gulf, the amount of suitable habitat available to right whales in any reproductive state remained relatively stable over the study period ( Fig. 3; Pearson's R = 0.15−0.40, all p > 0.05). However, the location of suitable habitat within the northern gulf varied between years (Fig. S5). For instance, suitable prey densities to support resting right whales in 2008 and 2009 were spread in patches from the estuary to Anticosti Island, and to the west of Newfoundland. In comparison, 2007, 2013 and 2014 had fewer grid cells showing suitable prey densities, which were mostly scattered along the north shore of the Gulf of St. Lawrence (north of Anticosti) and into the northeast gulf, west of Newfoundland, although sampling coverage in the northern gulf was more variable compared to in the south (Fig. S5a,b).
Similar spatio-temporal patterns in suitable habitat were found for pregnant females (Fig. S5c,d). However, there were little to no suitable prey densities for pregnant females under the suboptimal scenario in 2010, 2013 and 2015−2017, except for a few grid cells along the southern slope of the Laurentian Channel (2013,2016), and within the Strait of Belle Isle (2015) (Fig. S5c). For lactating females, predicted suitable habitat under optimal foraging conditions was largely concentrated in the southern gulf (2006Fig. S5f), with ≥ 50% of the water column containing suitable prey densities in the Shediac Valley area, around the Magdalen Islands, occasionally along the northwest tip of Prince Edward Island and south of Miramichi Bay. Only in 3 years (2006,2007,2011) were there sufficient prey densities to support the energy requirements of lactating females foraging under suboptimal conditions, albeit a very low number of grid cells. All of these sites were in the southern gulf (Fig. S5e). Composite maps of all years showed a broad area in the southern Gulf of St. Lawrence that had suitable prey densities (i.e. with grid cells having at least one 10 m depth bin with prey densities exceeding thresh-old requirements) to support res ting females in 6 or more years between 2006 and 2017 (Fig. 4). Areas with >10 years of consistently suitable prey densities for resting females also occurred at the entrance of Chaleur Bay and along the southern slope of the Laurentian Channel north of the Magdalen Shallows (Figs. 1 & 4). Similar patterns were found for pregnant and lactating females under optimal foraging conditions (Figs. 5 & 6). Areas showing 5−7 years of suitable prey densities for lactating females occurred off Miramichi Bay and the Shediac Valley region, and were spread across the Magdalen Shallows towards the Magdalen Islands.

DISCUSSION
We predicted locations of suitable foraging habitat for North Atlantic right whales within the Gulf of St. Lawrence by combining ecophysiological data (morphology, physiology, behavioural ecology, kinematics, and energetics) with 3-dimensional prey fields reflecting over a decade of temporal variability in prey densities and distributions. We found that the most suitable habitat for resting, pregnant and lactating right whales oc curred in the southern Gulf of St. Lawrence, which was consistent with the distribution of right whale sightings found during recent systematic surveys in the gulf (DFO 2019a(DFO , 2020, as well as with habitat predictions drawn from the spatial climatology of Calanus spp. biomass ).

Copepod biomass and distribution
Right whales are known to exploit energetically rich, diapausing latestage copepods during the summer (Baumgartner & Mate 2003). Diapausing Cala nus spp. in the North Atlantic can occur within a broad range of temperatures, salinities and densities (reviewed by Krumhansl et al. 2018), although they appear to favour depths with colder and denser waters, with temperatures < 5°C considered optimal for diapause (Saumweber & Durbin 2006). In the Gulf of St. Lawrence, the water column during the summer is highly stratified and includes a cold intermediate layer (<1°C), which is on average 50−100 m deep (Gilbert & Pettigrew 1997. In this region, the maximum abundance of diapausing C. finmarchicus and C. hyperboreus likely occurs in the deep channels and basins (such as the Laurentian Channel and Esquiman Channel) at depths that are routinely below the cold intermediate layer (Albouy-Boyer et al. 2016, Krumhansl et al. 2018).
In the southern Gulf of St. Lawrence, where densities of Calanus copepods suitable to support right whales were consistently predicted across the study period (2006−2017), the bathymetry is characterized by a broad, shallow shelf known as the Magdalen Shallows, which varies in depth from 60 to 80 m. C. finmarchicus and C. hyperboreus in this region are mainly transported from deeper regions (Zakardjian et al. 2003, Brennan et al. 2019, and are concentrated in shallower-than-normal waters during diapause, which compresses them against the sea floor (Kaartvedt 1996, Krumhansl et al. 2018). Under such conditions, maximum Calanus spp. biomass densities (g m −3 in 10 m depth strata) can be 5− 6 times greater at a 60 m deep station (southern gulf) than at a 300 m station in the deep Laurentian Channel . A combination of concentrated Calanus spp. layers and shorter transit times to access this food source may thus provide advantageous foraging conditions for right whales, and might also explain why such a broad area of the southern Gulf had suitable prey densities across multiple years.
The copepod biomass in the Gulf of St. Lawrence is predominantly C. hyperboreus (Sorochan et al. 2019), a species of unknown dietary contribution to right whales. This species declined in abundance from 2006 to 2017 in the southern Gulf of St. Lawrence and largely drove our conclusion that right whale habitat quality declined over time. Further north in the gulf, densities of C. hyperboreus remained relatively unchanged or were slightly higher during this period (Lehoux et al. 2020).
A similar conclusion about habitat quality for right whales in the gulf can be drawn from the densities of C. finmarchicus, which were low throughout the 2006−2017 period (Lehoux et al. 2020). Therefore, there is no clear indication that feeding conditions notably improved for right whales in the Gulf of St. Lawrence in the later years of our study, although they may have been better than in other foraging habitats. This aligns with the observation that C. fin- The southern gulf represents a population sink for Calanus spp. which must be re-populated in the spring of each year by transport of the new generation from the deep northern gulf (Runge et al. 1999, Za-kardjian et al. 2003, Blais et al. 2019, Brennan et al. 2019. Given that Calanus spp. abundance in the northern gulf did not substantially fluctuate between 2006− 2010(Blais et al. 2019, Lehoux et al. 2020, interannual changes in circulation patterns likely drove the decreased biomass of Calanus spp. ob served in the southern gulf . These changes in circulation patterns in the southern gulf likely stem from variations in local winds and freshwater runoff from the St. Lawrence River (and from other major tributaries) during spring.
The decline in Calanus spp. abundance in traditional feeding areas corresponds to a period of lower right whale calving rates and reduced use of these areas (Davis et al. 2017, Davies et al. 2019). Under such conditions, exploratory movements from a larger number of right whales seeking alternative feeding areas would be expected. Such exploratory behaviour may explain the increased acoustic detections of right whales in the Gulf of St. Lawrence since at least 2015 (Simard et al. 2019), as well as the recent abundance indices from systematic aerial surveys that suggest 40−50% of the right whale population visited the Gulf of St. Lawrence at some point in 2018 and 2019 (DFO 2020). Thus, the increased use of the Gulf of St. Lawrence by right whales in recent years does not necessarily mean they have found good habitat, given our finding that the availability of Calanus copepods in the gulf may be insufficient to allow for successful reproduction of North Atlantic right whales -a conclusion that is consistent with the continued low calving rates of this population.

Factors influencing right whale energy requirements
With their specialized ram filter-feeding strategy, right whales rely on the environment to concentrate their prey into energetically profitable layers. Density of prey organisms can thus be a key driver of foraging behaviour (Kraus & Rolland 2007). Right whales have been observed to cease foraging when prey densities fall below a certain range (Baumgartner & Mate 2003). These threshold prey densities appear to range from 800 to 4000 copepods m −3 , or 0.2−2.4 g m −3 depending on the average dry weights of different-sized C. finmarchicus C5 copepods (0.0002− 0.0006 g dry; Table 5). These minimum prey density estimates (recorded in the vicinity of foraging right whales in the Northwest Atlantic) are consistent with our findings (assuming 1 copepod weighs ~0.0006 g). Optimal prey density likely fluctuates with prey quality. In our case, we used a range of energy content values for late-stage C. finmarchicus and C. hyperbo reus to predict minimum prey densities for right whales. However, feeding on more or less nutritive prey sources throughout the year would influence our predictions for where suitable foraging habitat occurs for right whales.
Reductions in copepod size and energy content are likely to occur as environmental conditions continue to warm (e.g. Mc -Kinstry et al. 2013). This phenomenon ap pears to be underway in the Gulf of St. Lawrence, particularly in the southern gulf, where warmer sea temperatures correlate with reduced body sizes of C. fin marchicus during the summer months (when body sizes since 2013 are compared with the previous 2 decades; Sorochan et al. 2019). Feeding on smaller and less energetically rewarding copepods might mean that right whales would have to spend more time foraging, and more time finding areas that have higher-quality prey patches.
Changes in copepod community composition are expected to occur as the ocean warms, while copepod species that are typical of temperate latitudes are predicted to expand their range northward. This gradual change toward a more temperate copepod community is being observed on the Scotian Shelf and to a lesser extent in the Gulf of St. Lawrence (Blais et al. 2019, DFO 2019b. However, these new species of cope pods are unlikely to be a high-quality food for right whales because they are smaller and have a lower lipid content (see Lehoux et al. 2020). Moreover, the low filtering efficiency of right whale baleen to trap these smaller copepod species (unless they occur in exceptional densities) is likely a disadvantage to exploiting these new food sources (Mayo et al. 2001). Some of the densest euphausiid (krill) aggregations in the Gulf of St. Lawrence have been reported in the areas we identified as being suitable for North Atlantic right whales to forage (McQuinn et al. 2015, Plourde et al. 2016. However, there is limited evidence that right whales feed on euphausiids (except in the Antarctic where they occur in extreme densities, as reviewed by Wade et al. 2011; see also Tormosov et al. 1998), even though both euphausiids and copepods have been detected near feeding right whales in the North Atlantic and North Pacific, as well as in stomach contents (Wade et al. 2011). A study examining habitat suitability using the same bioenergetics model we used for North Atlantic right whales concluded that euphausiids are unlikely to be an important prey in the Gulf of St. Lawrence if right whales pursue them at their usual swim speeds (Lehoux et al. 2020). Right whales employ a foraging tactic (slow ram feeding) that is very different from krill-eating predators such as blue and fin whales (rapid-acceleration lunge feeding; Goldbogen et al. 2017). Right whales would need to swim at speeds approximately 5 knots faster than their usual speed to efficiently capture euphausiids (Hamner et al. 1988) -a behaviour undocumented in North Atlantic right whales (Murison & Gaskin 1989, Wade et al. 2011. Southern right whales, however, are able to effectively exploit Euphausia superba (Hamner et al. 1988, discussed by Harcourt et al. 2019) due perhaps to E. superba having different evasive and swarming behaviours, and/or higher biomass compared to krill species in the western North Atlantic (Tarling & Fielding 2016).
Fully determining the suitability of euphausiids as a food source for North Atlantic right whales will require assessing whether the energy they might gain from feeding on observed densities of krill can be balanced with the increased energy (associated with swimming at faster speeds) they would have to expend to capture it. Right whale dietary studies in the Gulf of St. Lawrence would also help support or refute our assumption that right whales feed primarily on late-stage Calanus spp., and are not adapted to consume euphausiids. Conceivably, additional foraging habitat could be exploited if right whales were to also opportunistically feed on krill (likely juvenile stages) in the Gulf of St. Lawrence. Bioenergetic modelling to help answer these questions can be further refined by incorporating information on prey species being targeted, as well as threshold, average and maximum prey densities sampled in the vicinity of foraging right whales on this feeding ground.
We chose to model the energy requirements of a 14 m adult female weighing 35 000 kg based on the morphometric findings of Moore et al. (2005), and used average calf mass at birth from Fortune et al. (2012) to estimate the cost of pregnancy. Our model did not account for growth in mass, which, although small, still exists for adult animals (Fortune et al. 2020b). Recently updated growth models for North Atlantic right whales using a broader dataset of necro psied individuals predicted shorter mean lengths (~13.5 m) for adults weighing 35 000 kg, and a slightly lower mean mass at birth (Fortune et al. 2020b). In performing a sensitivity analysis, we found that reducing adult body length by 0.5 m and reducing birth mass had little effect on energy expenditure estimates and had no effect on minimum prey density requirements, suggesting that our model is robust in capturing energy requirements for a greater range of adult lengths and calf masses.
Our mean estimates of annual energy expenditure for all 3 right whale reproductive states ( Beardsley et al. (1996)  . Adult male right whales require ~6% less energy than resting adult females due to differences in body composition and lower residency times on breeding grounds (Fortune et al. 2013). Although we did not include adult males in our study, their lower daily energy requirements would likely have resulted in more suitable foraging habitat being available to males than to adult females in the Gulf of St. Lawrence. In our model, we assumed that the energy requirement of a lactating female was a combination of her metabolism and that of her calf. Some lactating mothers may be unable to meet their daily energy requirements to arrive at zero or positive energy balance, resulting in reduced body condition that will take >1 recovery year for them to rebuild lipid stores and support another pregnancy. This means that prey densities that are 'good enough' to support calf growth and mother survival for 1 feeding season, may not cover the actual energy re quirements of lactation and may result in an energy deficit that mothers will pay at a later date (C. Miller et al. 2012, Christiansen et al. 2020. Unfortunately, the minimum daily energy intake needed to support a lactating female over the course of lactation, and the duration a negative energy balance (i.e. weight loss) could be sustained are unknown.
Females presumably have some plasticity in their reproductive behaviour to optimize their survival and that of their offspring. Females with calves may thus employ adaptive strategies that go beyond our model assumptions to cope with instability in foraging conditions. They may, for example, catabolize their energy stores to help offset any energetic im balance, extend time on feeding grounds to maxi mise food intake, reduce the number of hours spent foraging per day to reduce energy expenditure, forego migration to reduce travel costs, shorten or extend the lactation period depending on their provisioning strategy and/or extend their inter-calving interval to permit sufficient time to replenish their energy stores.
Other factors contributing to annual energy balance are daily and seasonal activity budgets, namely how much time is spent engaged in various behavioural states, and how this varies over the course of the year. The time−activity budget in our bioenergetics model (Table 3) was based on studies of right whale behaviour, migratory movements and residency times on wintering grounds in southeastern US waters. Gowan et al. (2019) showed that use of these wintering grounds varies considerably from year to year and across demographic groups, and that the probability of migration increases in years following higher C. finmarchicus abundance. Thus, reproductively mature females may choose to remain in northern latitudes during the fall, winter and spring, to maximise potential feeding opportunities and improve body condition.
In the absence of data on right whale activity budgets during the non-summer months in northern latitudes, or knowing what prey densities may be available, it is difficult to infer what effect continued oc cupancy of northern latitudes throughout the year would have on suitable habitat predictions in the Gulf of St. Lawrence. It would be informative to evaluate energy returns for whales that remain in northern latitudes and continue to feed in non-summer months given that foraging is more costly than travelling. We cannot dismiss the possibility that the overall daily energy requirements of right whales may be reduced in years when reproductive females choose to skip migration, potentially opening up more suitable foraging areas in the Gulf of St. Lawrence compared to our predictions. More information is therefore needed on right whale occurrence, residency and habitat use in northern latitudes such as the Gulf of St. Lawrence, Scotian Shelf and Newfoundland waters throughout the year (DFO 2019a, 2020).
Energy intake that our modelled right whales required partly depended on the number of hours per day and the number of days per year spent foraging. We assumed they spent a constant number of hours feeding per day (within a range of values) based on prior studies. In reality, however, whales may not feed for a day or more while looking for food, and may spend more hours per day feeding once a highquality food patch is found (e.g. Fortune et al. 2020a).
Without information on the daily activity budgets of right whales in the Gulf of St. Lawrence, we believe the range of daily foraging times we used from elsewhere in the right whale range were reasonable reflections of average times spent feeding over the season and were comparable to daily feeding times recorded for other baleen whales (Heide-Jørgensen et al. 2013). However, this assumption would benefit from validation with field observations. In the same vein, we assumed that right whales were physiologically capable of foraging at depths below maximum observed foraging depths in other habitats (ca. 130−140 m;Nousek McGregor 2010, Baumgartner et al. 2017. Evidence from bowhead whales Balaena mysticetus suggests that this is a plausible assumption. Bowhead whales share similar morphological and ecological characteristics with right whales, and have been observed performing U-and square-shaped dives (presumed foraging) to a maximum depth of 650 m, and V-shaped dives (presumed exploratory) to a maximum of 582 m (Heide-Jørgensen et al. 2013, Fortune et al. 2020a. Similar field studies are needed to document average and maximum depths of foraging dives in the Gulf of St. Lawrence, as well as determine how right whales use the water column throughout the season.

Comparing suitable habitat predictions to sightings data
Using the reproductive success of North Atlantic right whale females that visited the Gulf of St. Lawrence to validate our habitat suitability predictions is challenging because few systematic surveys were undertaken to locate and identify individual right whales in the Gulf of St. Lawrence until 2015. Unfortunately, the most extensive and sustained surveys were undertaken in this region following the die-off of multiple North Atlantic right whales in June 2017 (Daoust et al. 2017). Thus, there are insufficient sightings data to reliably compare the reproductive success of mature females that frequented the gulf with the predictions of our habitat suitability model (for the period 2006−2017).
Sightings data from 2017 and 2018 indicate that approximately one-third of reproductively mature female right whales (i.e. those that have calved at least once) visited the Gulf of St. Lawrence, albeit for an unknown duration (Table 6; , Right Whale Consortium 2020. Thus, despite the increase in right whale acoustic detections since 2015 (Simard et al. 2019), and the relatively high number of right whales identified during systematic surveys since 2017 (both suggesting an increase in occupancy of the Gulf of St. Lawrence by this population), the continued low calving rates suggest that increased use of the Gulf of St. Lawrence has not resulted in higher reproductive success in recent years. This is consistent with our conclusion that foraging conditions in the gulf were not favourable for right whale reproduction during the later years of our time series.
The fact that few females with calves have been observed within the gulf since 2016 ( Table 6) could imply that prey densities were still worth feeding on, but were insufficient to maintain net 0 or positive energy balance, which may explain the longer postlactation recovery periods and calving intervals (mean calving interval was 7 yr in 2019 compared to their historic 3 yr calving cycle; Knowlton et al. 1994. Another factor contributing to the lower calving rates and longer calving intervals of some females could be the added energetic burden of chronic entanglement in fishing gear -a plight which many individuals encounter (van der Hoop et al. 2017a). Conducting regular systematic surveys to determine the number, demography and residency period of right whales in the Gulf of St. Lawrence would help to assess these possibilities and further test the validity of our habitat suitability predictions.
In 2015 and 2016, survey effort in the GSL was considerably less than in subsequent years, making sightings data largely underestimated and unreliable. The proportion of reproductively mature females (resting, pregnant or lactating) identified in the GSL is therefore not shown for 2016 b In 2019, sightings data from the GSL were not completely processed at the time of writing. Total number of pregnant females in 2019 can be inferred from number of lactating females observed in 2020. The proportion of reproductively mature females (resting, pregnant or lactating) identified in the GSL is therefore not shown for 2019 Table 6. Total number of reproductively mature female North Atlantic right whales (i.e. those that calved at least once), total number of calves and total number of resting (Rest), pregnant (Preg) and lactating (Lact) females in the population from 2016 to 2019, as well as total number of mature females seen in the Gulf of St. Lawrence (GSL) in 2017−2018. Numbers of resting, pregnant, and lactating females in the GSL are shown both in absolute number and relative (%) to the total number of females in this reproductive state (in parentheses). Lactating females are those in repeatedly close association with a calf, and pregnant females are those seen with a calf the following year (and thus does not account for potential reproductive failure). NA: proportion invalid given the divider was zero. Data from  and Right Whale Consortium (2020)

Conclusions
Examining interannual variations of the 3dimensional Calanus spp. prey field in the Gulf of St. Lawrence (2006 relative to the energetic needs of reproductively mature female North Atlantic right whales indicates that prey densities should have been sufficient to support resting, pregnant and lactating females in most (≥ 6) years in the southern gulf (from Shediac Valley to the Magdalen Islands). However, we identified few other suitable foraging areas elsewhere in the gulf to support these 3 reproductive stages and noted a significant decline in habitat quality after 2014 (particularly for lactating females) due to reduced Calanus spp. biomass in the southern gulf.
As ocean temperatures continue to rise, negative trends in both the occurrence and abundance of C. finmarchicus are anticipated at the southern edge of its range (Reygondeau & Beaugrand 2011, Villarino et al. 2015, Grieve et al. 2017). In the Gulf of St. Lawrence, increasing bottom water temperatures ) might exceed thermal optima for C. finmarchicus and C. hyperboreus (Reygondeau & Beaugrand 2011), potentially compromising the suitability of this foraging area for North Atlantic right whales. Shelf waters off Newfoundland and Labrador might offer a more thermally stable habitat for C. finmarchicus compared to other areas (Loder & Wang 2015, Saba et al. 2016, where densities of C. finmarchicus have been comparable or higher than those reported for the Gulf of Maine and Gulf of St. Lawrence (Sorochan et al. 2019), and are likely to remain a habitat for this species in the future (Reygondeau & Beaugrand 2011).
Long-lived migratory megafauna such as whales that rely on memory of long-term average phenologies to locate quality food patches will need to adapt to altered distributions and abundances of key forage species (Abrahms et al. 2019). Varying patterns in North Atlantic right whale seasonal movements over the past decade are likely responses to shifts in prey distribution, emphasizing the need to identify areas of frequent use, where novel and dynamic management strategies should be implemented to reduce anthropogenic mortality and preserve areas that support vital functions.