Climate-Associated Regime Shifts Drive Decadal-Scale Variability in Recovery of North Atlantic Right Whale Population

USAGE Permission is granted to copy this article for use in teaching and research. Republication, systematic reproduction, or collective redistribution of any portion of this article by photocopy machine, reposting, or other means is permitted only with the approval of The Oceanography Society. Send all correspondence to: info@tos.org or The Oceanography Society, PO Box 1931, Rockville, MD 20849-1931, USA. Oceanography THE OFFICIAL MAGAZINE OF THE OCEANOGRAPHY SOCIETY


INTRODUCTION
Humans began hunting the North Atlantic right whale (Eubalaena glacialis) nearly a millennium ago.By the end of the nineteenth century, the population was so depleted that it was of little commercial value to the whaling industry (Allen, 1908).In 1935, the right whale population first received protected status from the League of Nations, and, since 1949, a complete moratorium on hunting has been in place and overseen by the International Whaling Commission (Best et al., 2001).Despite the end of commercial whaling in the mid-twentieth century, the recovery of this endangered population has been gradual and highly variable.Accurate demographic studies only became possible after an extensive and ongoing effort was initiated during the 1970s to photograph and catalog all individuals in the population (Kraus et al., 1986).Subsequent analyses of the demographic data indicated that the population's growth rate increased gradually during the 1980s, but then declined sharply during the 1990s (Fujiwara and Caswell, 2001).Demographic projections based on data from the early1990s suggested that the population was on a trajectory that would lead to its eventual extinction in less than 200 years (Fujiwara and Caswell, 2001).
As news of these demographic projections spread within the right whale research community, a consensus view emerged that the population would continue to decline unless the right whale's elevated mortality rates associated with ship strikes and entanglement in fishing gear could be significantly reduced (Fujiwara and Caswell, 2001;Kareiva, 2001;Waring et al., 2012).In contrast to this expectation, the right whale population began to recover during the following decade.Despite high mortality rates and even more dire demographic projections during the first decade of the 2000s (Kraus et al., 2005), the population grew from ~ 340 animals at the beginning of the decade to ~ 486 animals by 2010 (Figure 1a).A major factor in this recovery was the 128% increase in female-specific average annual calf production between 2001 and 2010 relative to the previous decade (Figure 1b).Here, we use a data-driven, stochastic reproduction model to explore the ecological underpinnings for this dramatic increase in right whale reproduction.

Right Whale Population Data
North Atlantic right whales have been photographically cataloged in a consistent manner since 1980 and are identified using unique markings, scars, and callosities (rough patches of tissue found on the animals' heads) (Kraus et al., 1986).Population growth rates reported in this paper use what are considered to be the best estimates of the total population and numbers of reproductively viable female whales.The North Atlantic Right Whale Consortium (http://www.narwc.org)provided us with a list of all known whales along with the years in which they died or were last sighted.
Six years missing is the standard used by the Consortium to define a whale that is presumed to have died.Annual estimates of the total number of whales presumed to be alive therefore include any whales on the list that are not known to have died or have not gone missing for six or more years without a subsequent sighting.Because of the potential for bias in using this six-year rule, we only used population data up until 2007 in our models.Among whales presumed to be alive, females considered reproductively viable are those known to have given birth or to have reached nine years of age, the average age of first parturition (Hamilton et al., 1998).Whales categorized as senescent are also removed from the time series of reproductively viable females (Knowlton et al., 1994).region, mother/calf pairs have an especially high sighting probability; therefore, it is assumed that all newborn calves have been observed.

Right Whale Reproduction Model
The  The two probabilities optimized in this study are Ø 21 and Ø 32 , which represent the probability of a female transitioning from the recovery state (1) to the pregnant state (2), and the probability of a female transitioning from the pregnant state (2) to the nursing state (3), respectively.The transitional probabilities were estimated as logistic functions to constrain the probabilities between 0 and 1 while offering flexibility in the shape of the function: (2) 1 + e β32 * X e β32 * X where the vector notations β 21 * X and β 32 * X each represent a linear combination of an intercept and coefficient(s) multiplied by the independent prey variable(s) X.These transitional probabilities were fit into a demographic matrix model, and the parameter vectors β 21 and β 32 were estimated to yield a predicted annual calf production time series most closely resembling the observed time series.
The model estimates of β 21 and β 32 predict different functional responses of the transitional probability Ø 21 and the transitional probability Ø 32 as functions of prey abundance.Ø 21 increases gradually with increasing values of the Calanus finmarchicus abundance index, yielding a relatively linear relationship.However, Ø 32 behaves like a quasi-step function, with the transition to nursing a calf changing abruptly from highly improbable to highly probable over a narrow range of C. finmarchicus abundance values.This abrupt transition occurs at abundance values slightly below the climatological average for that bimonthly time period.
Atlantic (Greene et al., 2008;MERCINA Working Group, 2012).In the Gulf of Maine, the low-salinity waters altered the timing and extent of water-column stratification, which subsequently impacted the production and seasonal cycles of phytoplankton, zooplankton, and higher-trophic-level consumers in the ecosystem (Greene and Pershing, 2007;Greene et al., 2008;MERCINA Working Group, 2012) Figure 1.Time series of North Atlantic right whale population size from 1980 to 2010.(a) Total number of all whales (dark blue), number of reproductively viable females (red).(b) Number of calf births (light blue).
Ø 21 ) or become pregnant (Ø 21 ).Pregnant females can give birth and enter the nursing state (Ø 32 ) or abort the pregnancy and reenter the recovering state (1 -Ø 32 ).Nursing females can only transition to the recovering state (Ø 13 = 1).Using this sequence of three reproductive states, we constructed the following transitional probability matrix: ij in the matrix is the probability of a reproductively viable female transitioning from state j to state i in a year.Projection matrix A is multiplied by the female abundance vector N t-1, or the number of living viable females in each of the three reproductive states during year t-1, to estimate the female abundance vector N t during the following year t:N t = [A] * N t-1The two probabilities optimized in this study are Ø 21 and Ø 32 , which represent the probability of a female transitioning from the recovering state (1) to the pregnant state (2), and the probability of a female transitioning from the pregnant state (2) to the nursing state (3), respectively.Transitional probabilities were estimated as logistic functions dependent on C. finmarchicus abundance.Model parameter vectors were optimized to yield a predicted calf production time series that best fits the observed time series provided by the North Atlantic Right Whale Consortium (Box 1).Calanus finmarchicus abundance indices were estimated from Gulf of Maine Continuous Plankton Recorder (CPR) survey data (Greene et al., 2013) collected from 1980 to 2007.Despite the sampling limitations of the CPR and the averaging out of spatial and temporal patchiness, this index has proven to be a remarkably useful proxy for characterizing interannual to interdecadal variability in C. finmarchicus abundance (MERCINA Working Group, 2001, 2004; Greene et al., 2008).Six bimonthly C. finmarchicus abundance indices were determined for the entire Gulf of Maine region and for each of four geographical subregions: Massachusetts Bay (MB), Western Gulf of Maine (WGOM), Eastern Gulf of Maine (EGOM), and Scotian Shelf (SS) (Figure 2b).An annual average C. finmarchicus abundance index was also determined for the entire Gulf of Maine region and for each subregion.Combinations of all indices were added and evaluated in a stepwise fashion to determine the best overall model fit to the annual calf production time series.RE SULTS Annual calf production estimates from our model demonstrate the tight coupling between right whale reproduction and prey abundance over the past three decades (Figure 3a,b).Results from the model incorporating bimonthly and regional variations in prey abundance (Figure 3b) fit the observed data better than results from a temporally and spatially averaged version of the model Erin L. Meyer-Gutbrod (elg82@ cornell.edu) is a PhD candidate, and Charles H. Greene (chg2@cornell.edu) is Director and Professor, both in the Ocean Resources and Ecosystems Program, Earth and Atmospheric Sciences, Cornell University, Ithaca, NY, USA.

Figure 2 .(
Figure 2. (a) North Atlantic right whale reproduction model.Reproductive females can be (1) in the recovering state between pregnancies, (2) pregnant, or (3) in the nursing state.Transitional probabilities between states are determined as functions of Calanus finmarchicus abundance in the Gulf of Maine.(b) Gulf of Maine Continuous Plankton Recorder (CPR) survey area used to characterize prey availability to right whales.The CPR survey area is divided into four geographical subregions: Massachusetts Bay (MB), Western Gulf of Maine (WGOM), Eastern Gulf of Maine (EGOM), and Scotian Shelf (SS).CPR survey sampling (black dots) in the Gulf of Maine occurs at approximately monthly intervals.

Figure 3 .
Figure 3. (a).Time series of C. finmarchicus annual abundance index (Greene et al., 2013) estimated for the entire Gulf of Maine CPR survey area.(b,c) Time series of annual calf production (calves • year -1 ) observed (black lines in b and c) and predicted by a model driven by bimonthly-and geographic-specific abundance estimates of C. finmarchicus in the Gulf of Maine (red line in b) and predicted by a model driven by annual abundance estimates of C. finmarchicus averaged for entire Gulf of Maine (blue line in c).The pale red and blue shading surrounding the model predictions correspond to the 95% confidence intervals.

Figure 4 .
Figure 4. Time series of female-specific annual calf production (calves • 100 females -1 • year -1 ).Decadal averages of female-specific annual calf production are shown with blue lines.