Long-term population size of the North Atlantic humpback whale within the context of worldwide population structure
- First Online:
- Cite this article as:
- Ruegg, K., Rosenbaum, H.C., Anderson, E.C. et al. Conserv Genet (2013) 14: 103. doi:10.1007/s10592-012-0432-0
Once hunted to the brink of extinction, humpback whales (Megaptera novaeangliae) in the North Atlantic have recently been increasing in numbers. However, uncertain information on past abundance makes it difficult to assess the extent of the recovery in this species. While estimates of pre-exploitation abundance based upon catch data suggest the population might be approaching pre-whaling numbers, estimates based on mtDNA genetic diversity suggest they are still only a fraction of their past abundance levels. The difference between the two estimates could be accounted for by inaccuracies in the catch record, by uncertainties surrounding the genetic estimate, or by differences in the timescale to which the two estimates apply. Here we report an estimate of long-term population size based on nuclear gene diversity. We increase the reliability of our genetic estimate by increasing the number of loci, incorporating uncertainty in each parameter and increasing sampling across the geographic range. We report an estimate of long-term population size in the North Atlantic humpback of ~112,000 individuals (95 % CI 45,000–235,000). This value is 2–3 fold higher than estimates based upon catch data. This persistent difference between estimates parallels difficulties encountered by population models in explaining the historical crash of North Atlantic humpback whales. The remaining discrepancy between genetic and catch-record values, and the failure of population models, highlights a need for continued evaluation of whale population growth and shifts over time, and continued caution about changing the conservation status of this population.
KeywordsEffective population sizeHumpback whaleCensus population sizePopulation structure
Over-exploitation has resulted in the collapse of many marine populations (Pauly et al. 1998; Myers and Worm 2003; Estes et al. 2006). In some cases, however, national or international protection has led to the recovery of previously threatened or endangered species (reviewed within Scott et al. 2005). Humpback whales (Megaptera novaengliae) in the North Atlantic were severely depleted as a result of intense hunting during the 19th and 20th centuries (Mitchell and Reeves 1983; Braham 1984; Winn and Reichley 1985) and are currently listed as ‘endangered’ or ‘vulnerable’ by various governments and international conservation organizations (Klinowska 1991). Before the International Whaling Commission (IWC) banned commercial whaling in the North Atlantic in 1955, it was estimated that this population was reduced to <1,000 individuals (Mitchell and Reeves 1983; Katona and Beard 1990). After many decades of legal protection, humpback whales have increased in numbers (Stevick et al. 2003) and recent survey estimates suggest that they may be approaching 20,000 animals (Smith and Pike 2009). Such increases in population size within the North Atlantic have led the IUCN and the US to re-evaluate their conservation status.
Assessing the recovery of previously depleted populations requires knowledge of past population sizes, but robust estimates of past abundance can be difficult to attain. Different approaches to estimating pre-whaling population sizes can lead to starkly different conclusions about the extent of recovery in the North Atlantic humpback whale (Roman and Palumbi 2003; Holt and Mitchell 2004; Punt et al. 2006). Traditionally, the IWC has relied upon population dynamic models that use a combination of information on current abundance, catch records, rates of increase, and population structure to estimate changes in population size through time (Punt et al. 2006). Recent model estimates for the North Atlantic humpback whale suggest a pre-whaling population size of between 20,000 and 46,000 individuals, depending upon the catch data used (Punt et al. 2006). Given current abundance estimates of ~17,700 individuals (Smith and Pike 2009), population model-based estimates suggest that humpbacks in the North Atlantic are approaching the lower boundary of their pre-whaling numbers. Alternatively, genetic-based estimates of pre-whaling abundance use the relationship between genetic diversity (θ) and effective population size (Ne) (θ = 4Νeμ, where μ is the average mutation rate) to estimate the long-term population size of North Atlantic Humpback whales (Roman and Palumbi 2003). Genetic estimates calculated using mitochondrial DNA (mtDNA) control region sequence, suggest a pre-whaling abundance of 150,000–240,000 depending upon the mutation rate employed (Roman and Palumbi 2003; Alter and Palumbi 2009). These genetic estimates suggest that there were substantially more whales prior to whaling than previously believed.
The discrepancy between estimates of pre-whaling abundance based upon catch records and estimates based upon genetic variability has been the subject of vigorous debate (Lubick 2003; Holt and Mitchell 2004; Clapham et al. 2005). Some argue that unavoidable uncertainties in the catch record may have led to underestimates in the number of whales removed from the North Atlantic due to whaling (Palumbi and Roman 2007). However, a recent review and re-reading of whaling records revealed only slight increases in the numbers of North Atlantic humpback whales estimated to be killed as result of whaling (from 29,000 to 30,852 total catches) (IWC 2002, 2003; Smith and Reeves 2010). Others argue that genetic estimates of long-term abundance may be inaccurate as a result of reliance on a single locus, uncertainty surrounding mutation rates and generation times, the potential influence of incomplete sampling, and the evolutionary time-scale to which a genetic estimate applies (Lubick 2003; Holt and Mitchell 2004; Clapham et al. 2005). While recent estimates of long-term population size in gray and minke whales have reduced some of these uncertainties through a variety of methodological improvements (Alter et al. 2007; Alter and Palumbi 2009; Ruegg et al. 2010), humpback whales are particularly challenging because of their complex oceanic and worldwide population structure (Baker et al. 1993; Palsboll et al. 1995; Rosenbaum et al. 2009).
Population structure may affect estimates of long-term effective population size (Ne) in a variety of ways depending upon the extent of isolation between populations. Theoretical models suggest meta-population Ne and sub-population Ne converge as migration between sub-populations increases (Hudson 1991; Waples 2010). It has also been shown that even low migration rates between sub-populations can cause an estimate of long-term Ne that is based upon samples from one sub-population to approximate the long-term Ne of the whole meta-population population (Hudson 1991). Thus, if there is migration between sub-populations, then the absence of samples from one sub-population should have very little effect on an estimate of long-term Ne for the whole meta-population because the two values will be equivalent. Alternatively, if there is no migration between populations and one uses samples from an isolated sub-population, then an estimate of long-term Ne for the meta-population will be downwardly biased. Thus, in order to identify the impact of population structure on Ne it is important to also measure levels of population structure.
Presently, humpback whales are divided into three oceanic populations, the North Atlantic, the North Pacific and the Southern Hemisphere, based on genetic and tagging data suggesting limited migration between ocean basins (Mackintosh 1965; Baker et al. 1993). Previous analysis of worldwide population structure based upon mtDNA suggests that humpback whales from the North Atlantic are most strongly differentiated from those in the North Pacific and less strongly differentiated from those in the Southern Hemisphere (Table 2, Baker et al. 1993). Strong divergence between North Atlantic and North Pacific humpback whales is thought to result from the fact that sea ice has likely blocked the main northern migratory corridor between the two groups since the Sangamonian Interglacial period (~140,000 years ago). As a result, genetic diversity within the North Atlantic is unlikely to be strongly influenced by past migration from the North Pacific. Thus, while we will test the assumption that gene flow with the North Pacific does not influence Ne in the North Atlantic, the main focus of our analysis will be on populations from the North Atlantic and the Southern Hemisphere.
Patterns of migratory connectivity in the Southern Hemisphere are less well understood, but recent evidence based upon mtDNA suggests low, but significant sub-population structure between Southwestern Atlantic, Southeastern Atlantic, and Southwestern Indian Ocean groups (Breeding Stocks A, B, and C respectively; Fst range 0.0029–0.0166) with the Northern Indian Ocean (stock X) falling out as strongly differentiated from all other groups (Fst range from 0.0797 to 0.1473) (Rosenbaum et al. 2009). In this study, we will use multiple nuclear loci and increased sampling from within each ocean basin to gain a better perspective on the impact of population structure on long-term population size in the North Atlantic.
We calculate the long-term population size of the North Atlantic humpback whale within the context of the worldwide population structure. We focus on an in-depth analysis of the North Atlantic and the Southern Hemisphere, with particular focus on the South Atlantic, because previous data indicate that these two populations are the most likely to have exchanged migrants during a time period that may impact an estimate of effective population size. We identify strongly differentiated populations within and between the North Atlantic and Southern Hemisphere using nine nuclear loci and a multi-locus genetic clustering method. We then estimate long-term population size of the North Atlantic humpback whale, while accounting for the possibility of migration with other strongly differentiated groups. Our new estimate of long-term population size in the North Atlantic is compared with previous genetic (mtDNA) and catch-based estimates in order to highlight remaining uncertainties in estimates of pre-whaling abundance and discuss important areas for future research.
Sample collection and sequencing
Genetic samples representing 173 individuals were collected from humpback whales across the Southern Hemisphere (South Atlantic and Indian Oceans) and the North Atlantic Ocean (for regional sample sizes see Fig. 1). Biopsy samples from living whales were collected with appropriate national permits using protocols approved by the American Museum of Natural History and the Oregon State University’s Animal Care and Use Committees. Samples were preserved in 70 % ethanol or salt saturated 20 % dimethyl sulfoxide solution (DMSO) and later stored at −20 °C until processed. Total genomic DNA was extracted using a standard phenol/chloroform extraction method or using a DNAeasy tissue kit (Quiagen).
Summary statistics for 9 introns sequenced in North Atlantic and Southern Hemisphere humpback whales
PHASE 2.1 (Stephens et al. 2001) was used to reconstruct gametic phase, defined as the original allele combination that an individual received from each of its parents, using a burn in of 10,000 iterations and a run length of 10,000 iterations. Using Arlequin ver. 3.0 (Excoffier et al. 2005) we found no significant linkage disequilibrium among loci after correcting for multiple comparisons. To determine if our sequences were evolving in a manner consistent with equilibrium and neutrality, Tajima (1989) and Fu (1997) tests were preformed using DnaSP (Rozas et al. 2003). In neutrally evolving sequences, both values will be approximately equal to zero, while balancing selection or population expansion will result in values that are significantly greater or less than zero, respectively. We also used DnaSP to calculate the minimum number of recombination events in the sample (Hudson and Kaplan 1985) and found that 3 of 9 loci showed evidence of recombination. As a result, in loci with evidence for recombination, coalescent simulations (n = 1,000) incorporating the per gene recombination parameter (R) were used to generate 95 % confidence intervals (CI) for both Tajima’s D and Fu’s Fs statistics.
Testing for population structure
An analysis of population structure was performed to investigate whether or not the major divisions within and between the North Atlantic and the Southern Hemisphere humpback whales remained with increased sampling in both regions. In order to properly account for migration that may impact our estimate of genetic variation (theta, θ) for the North Atlantic humpback whale, we test 3 population structure scenarios: (1) populations with no migration over recent evolutionary history (i.e. 4 Ne generations), (2) genetically distinct populations connected by very limited migration, and (3) sub-populations that may be biologically meaningful, but are exchanging migrants at a high enough rate that they cannot be distinguished using multi-locus clustering methods. To estimate long-term population size, sub-populations (scenario 3) were lumped into respective population categories (scenario 2) and populations with no possibility of migration with the North Atlantic over recent evolutionary history were considered separately (scenario 1).
Pairwise Fst within and between ocean basins at each locus as well as across all loci were calculated using the program Arlequin ver. 3.0 (Excoffier et al. 2005). A null distribution of Fst was generated through 1,000 permutations of the haplotypes between populations and the p value represents the proportion of permutations leading to an Fst larger than or equal to the observed value. To assess the potential for within and between ocean basin population structure within a multi-locus framework, we used the program Structure ver. 2.2 (Pritchard et al. 2000). Preliminary runs indicated that the power for assigning individuals to clusters dropped off significantly when individuals had missing data for more than 2 of the 9 loci. Therefore, individuals with missing data for more then 2 loci were removed from the multi-locus analysis in order to ensure that there was sufficient statistical power for assignment of all individuals to clusters. Structure requires unlinked markers, so the maximum a posteriori haplotypes from PHASE at each locus were recoded as alleles. We performed 3 independent runs at each K value (K = 1–5) using a burn-in period of 100,000 iterations and a run length of 500,000. The structure analysis was run using the admixture model with correlated allele frequencies with and without the location prior.
The location prior is intended to use location information to help identify more subtle population structure, without detecting structure that is not present (Hubisz et al. 2009) and we implement it here in an attempt to identify a signature of population structure in our data that may influence our subsequent estimate of q. Locations included the Gulf of Maine (GOM), Dominican Republic (DR), Gabon (GA), Brazil (BR) and Madagascar (BA) (Fig. 1, SI Table 1). Individuals from NF were grouped within the GOM location due to low sample size from NF, geographic proximity between NF and GOM and the lack of significant Fst values between NF and GOM (see “Results” section). We determined support for the number of clusters (K) by plotting the average ln [P(X|K)] of each model as a function of K and using the ad hoc ΔK statistic proposed by Evanno et al. (2005).
Using our knowledge of population structure, we employed genetic models (Kuhner 2006) that estimate long-term Ne while explicitly accounting for the possibility of migration between populations deemed distinct according to multi-locus clustering methods. We use LAMARC ver. 2.1.3 to simultaneously estimate θ while incorporating recombination and migration between ocean regions into the model. In contrast to summary statistic estimates of θ (θs, θπ, etc.), LAMARC accounts for uncertainty in the data by integrating over the space of possible genealogies using a Markov chain Monte Carlo (MCMC) procedure. In order to account for uncertainty in the data resulting from unknown gametic phase and to accommodate inter-locus variation in mutation rate, we followed the methods described in Ruegg et al. (2010). In short, to account for unknown gametic phase, LAMARC was run on 15 realizations from PHASE’s posterior distribution for each of 9 introns. In addition, as recommended by the LAMARC manual, we subsampled our data to restrict the input size for each LAMARC run to 20 sequences from each major population. Thus, for each of the 15 realizations from PHASE’s posterior, LAMARC was run on a different subsample of 10 randomly chosen individuals from each population. The final result from these 15 LAMARC runs was obtained by catenating the summaries from all the runs following the recommendations in the LAMARC manual for “poor man’s parallelization.” (Initially we attempted 3 random subsamples from each phasing—45 total LAMARC runs—but this exceeded the memory available to LAMARC). To accommodate interlocus variation in mutation rate, we implemented the gamma model for mutation rate variation within a Bayesian framework using an extension of the LAMARC package known of as GUFBUL (Gamma Updating for Bayesians Using LAMARC; Ruegg et al. 2010).
Our main objective was to estimate θ in the North Atlantic while accounting for the possibility of migration with the Southern Hemisphere. To this end we used a 2-population migration model in LAMARC on the full dataset that included nine loci (Table 1). The fact that ice has blocked the main northern migratory corridor between the North Pacific and the North Atlantic since the Sangamonian Interglacial period (~140,000 years ago) makes gene flow between the two populations unlikely. However, to further investigate the possibility that genetic diversity in the North Atlantic is influenced by migration with the North Pacific, we ran a LAMARC analysis using a 3-population migration matrix on 6 of the nine loci for which we had sequence data. θ values generated using the 3-population model were compared to values calculated using the 2-population model for each of the 6 loci.
Calculating census population size from θ
The conversion of θ into effective population size (Ne) is based upon the relationship θ = 4Neμ where μ is the average mutation rate. To calculate an average μ for North Atlantic Humpback whales, and to estimate uncertainty surrounding our estimate, we followed the methods described in Ruegg et al. (2010). In short, we sampled with replacement from among 9 previously published individual locus mutation rates for humpback whales; 1 of the individual locus mutation rates (PLP) was from Alter et al. (2007), while the remaining 8 were taken from a Bayesian analysis of baleen whale phylogeny and fossil history (Jackson et al. 2009). For each re-sampled locus, a sample mutation rate was drawn from the posterior distribution of the estimated mutation rate or, for PLP, uniformly from the 95 % confidence intervals on the mutation rate. This was repeated 9 times for each bootstrap replicate, and we performed 100,000 bootstrap replicates. The mean μ and the variability around that mean was obtained from these bootstrap replicates. To convert μ from units of mutations per base pair per year into mutations per base pair per generation requires an estimate of the generation length. To approximate generation length we sampled uniformly from within a range of possible values for North Atlantic humpback whales of between 12 and 24 years (Chittleborough 1965; Roman and Palumbi 2003; Taylor et al. 2007). While this lower bound on generation time, taken from Chittleborough’s (1965) estimate, may be low because of age-estimate inaccuracies, it is similar to the 14.5 year estimate for modern humpback whales from Taylor et al. (2007). Here we maintain the 12–24 year range in order to stay consistent with previous estimates of long-term population size in the North Atlantic humpback whale (Roman and Palumbi 2003), and discuss the implications of different generation times on estimates of Ne.
To convert Ne to census population size (Nc) requires an estimate of the ratio of mature adults to the effective number of adults (Nmature/Ne) and the proportion of juveniles in the population. Although Nmature/Ne is difficult to calculate in most natural populations, theory suggests this ratio approaches 2 in most populations with constant size (Nunney and Elam 1994). We based our estimate of Nmature/Ne on equation (1) in Nunney and Elam (1994): Ne = N/(2−T−1), where T = generation length. To approximate juvenile abundance we used catch and survey data to calculate (no. of adults + juveniles)/(no. adults) (Chittleborough 1965; Roman and Palumbi 2003). To incorporate uncertainty in juvenile abundance we sampled uniformly from within a range of likely values for North Atlantic humpback whales.
Tests for neutrality and equilibrium
Among the 9 nuclear introns, nucleotide diversity averaged 0.0014 (range 0.0002–0.0054), with an average of 5 haplotypes per locus (range 2–11) and an average of 43 samples from the North Atlantic and 101 samples from the Southern Hemisphere (Table 1). These values were similar to other baleen whale species for which data are available (gray whales: range 0.0031–0.00016; Alter et al. 2007). While Tajima’s D for ACT and Fu’s FS for ESD were significantly different from the simulated null distribution given p < 0.05, neither remained significant after Bonferroni correction for multiple comparison (corrected p = 0.05/18 tests = 0.003). The results of the Tajima’s D and Fu’s FS tests suggest the loci are evolving in a manner consistent with neutrality and equilibrium (Table 1).
Across the 9 loci, Fst ranged from 0 to 0.36 (SI Table 1), with 69 % (18 of 26) of the significant pairwise Fst values being between North Atlantic and Southern Hemisphere populations, 31 % of the significant comparisons being between populations in the Southern Hemisphere, and 0 % coming from comparisons between populations in the North Atlantic. When the North Atlantic and Southern Hemisphere populations were grouped into two groups, the overall Fst across all loci was 0.14. For the multi-locus analysis of population structure, inspection of the average log probability of the data (ln [P(X|K)]) and the ad hoc ΔK statistic of Evanno et al. (2005) indicated K = 2 was the most likely number of clusters in the data (SI Fig. 1a and b). A plot of the average ln [P(X|K)] of each model as a function of K showed the likelihood increased substantially with an increase in K from 1 to 2, but increased to a lesser extent or decreased thereafter (SI Fig. 1a). Similarly, ΔK was substantially greater for a K of 2 than for any other value of K (SI Fig. 1b). The results were the same without using the location prior (results not shown).
Estimating genetic diversity (θ)
Theta values for the North Atlantic estimated using a two population (NA and SH) migration matrix
Estimate of census population size from θ
To improve estimates of long-term population size in the North Atlantic humpback whale, we have addressed recommendations for larger numbers of genetic loci, a better perspective on the impact of population structure, greater confidence in the mutation rate, and a greater focus on the historical timeframe of genetic population estimates (Clapham et al. 2005). Our new estimate of long-term population size of ~112,000 individuals (95 % CI 45,000–235,000) is less than half of the previous mtDNA-based estimate of ~240,000 (95 % CI 156,000–401,000) (Roman and Palumbi 2003), but is very similar to a revised population number of 150,000 (95 % CI 45,000–180,000) based on a more accurate estimate of the mutation rate (Alter and Palumbi 2009). However, the median of our most recent estimates remains far higher than the highest pre-whaling abundance estimate based upon catch data (notional upper limit: 40,000–47,000) (Smith and Pike 2009) and the discrepancy between the estimates warrants further discussion.
Because genetic diversity within populations is strongly influenced by migration between populations, estimates of long-term population size must account for population structure.
We re-evaluated population structure within the humpback whale based upon previous work (Baker et al. 1993; Valsecchi et al. 1997; Olavarria et al. 2007; Rosenbaum et al. 2009) and our own multi-locus analysis. Our results confirm that θ in the North Atlantic has not been significantly influenced by migration with the North Pacific (SI Fig. 2), unlike recent reports for Bowhead whales (Alter et al. 2012). Thus, our main analysis focused upon the North Atlantic and the Southern Hemisphere. While the locus-by-locus analysis revealed some signal of sub-population structure within the Southern Hemisphere (SI Table 1), the results of our within ocean basin multi-locus analysis indicate a lack of significant population structure overall, even when a strong prior for the presence of multiple sub-populations was included (Hubisz et al. 2009) (Fig. 2a). Overall, both the analysis of the average log probability of the data and the ad hoc ΔK statistic indicate that K = 2 is the most likely number of clusters in the data (SI Fig. 1). Consistent with previous research (Valsecchi et al. 1997; Olavarria et al. 2007; Rosenbaum et al. 2009), our results suggest that humpback whales within each ocean basin consist of two distinct populations connected by some migration.
One limitation with our study was the lack of samples from the eastern North Atlantic where previous research suggests the existence of a genetically distinct sub-population (Valsecchi et al. 1997). It is possible that additional samples from this region may have increased the number of distinct clusters found within the North Atlantic. However, because even small amounts of migration will cause sub-population Ne and whole population Ne to converge (Waples 2010; Hudson 1991), the absence of samples from the eastern North Atlantic sub-population is not likely to have influenced our estimate of long-term Ne. If, contrary to previous research, there is no migration between eastern and western North Atlantic feeding groups, then including samples from the eastern North Atlantic would increase our estimate of long-term population size.
Attaining accurate estimates of mutation rates is a challenge common to all studies that use genetics to infer past population process (Ho et al. 2005; Emerson 2007). The difference between the original mtDNA-based estimate of ~240,000 (Roman and Palumbi 2003), the updated mtDNA-based estimate of ~150,000 (Alter and Palumbi 2009) and our multi-locus estimate of ~112,000 individuals highlights the importance of mutation rates to estimates of long-term population size. In their revised estimate, Alter and Palumbi (2009) recalibrated the control region mutation rate used in Roman and Palumbi (2003) by implementing a cytochrome b clock. Their analysis suggested that the previous mutation rate estimate was low by about two-fold because of multiple substitutions in the quickly evolving mtDNA control region. When the re-calibrated control region mutation rate is employed, the mtDNA-based estimate becomes statistically indistinguishable from the multi-locus estimate of long-term population size.
Here we estimate an average mutation rate across nine nuclear loci using a phylogenetic reconstruction of the baleen phylogeny and fossil history (Jackson et al. 2009). One advantage of our multi-locus nuclear estimate is that whale nuclear DNA has far less saturation of substitutions than the mtDNA control region, and thus is far less likely to be subject to the same rate problems. Furthermore, our multi-locus approach incorporates uncertainty that results from random variation in the coalescent history of each individual locus (Rosenberg and Nordborg 2002). To adequately reflect the uncertainty in mutation rates in our final estimate of long-term population size, we bootstrap resampled across the variation in individual locus mutation rates. Thus, the multi-locus nuclear estimate that we present here should be a more robust approximation of the long-term Ne than the preceding mtDNA-only estimates.
Uncertainty surrounding generation lengths interacts with mutation rate to determine estimates of long-term population size. Here we use a wide estimate of generation length for humpback whales ranging from 12 to 24 years (Chittleborough 1965; Roman and Palumbi 2003; Taylor et al. 2007) in order to remain consistent with previous estimates of long-term population size (Roman and Palumbi 2003). However, generation time in whales remains uncertain. Our lower bound of 12 years taken from Chittleborough (1965) is based on female age-size estimates from baleen condition, earplug layers and ovarian cycles. While Chittleborough (1965) provides the most extensive empirical data from which to estimate generation length in humpback whales, his estimates suffer from questions about age estimation (Gabriele et al. 2009; Best 2011) and whether older animals had already been culled (both of which would decrease estimates of generation length).
Taylor et al. (2007) estimated generation length for 58 cetacean species, including humpback whales, using mathematical models based on age at first reproduction and survival. They used an annual adult survival of 96 % and a first breeding age of 6 years to estimate a current generation time of 14.5 years and a stable pre-exploitation generation time of 21.5 years. While both of these estimates fall within our wide range on generation time, the similarity between the current generation time of 14.5 years and the Chittleborough (1965) estimate of 12 years further highlights how exploitation may skew age patterns towards younger individuals. Furthermore, the absence of empirical data makes model-based estimates such as these especially sensitive to underlying assumptions. Better estimates should come from age distributions of real, unexploited populations, but such data is not readily available.
A longer estimate of humpback whale generation length would decrease our estimate of long-term population size (because the mutation rate per generation would increase). For example, if we use the estimate of 21.5 years taken from Taylor et al. (2007) we would decrease our estimate of population size to ~90,000, bringing it closer to catch based estimates of pre-whaling abundance. If we use the full range of generation times estimated through models by Taylor et al. (2007) for Baleanopterid whales (18–31 years), our mean estimate of long term population size would be ~81,000 whales (95 % CI 34,000–163,000). These results highlight the sensitivity of genetic estimates of long-term population size to estimates of generation length. However, generation time would need to be much longer than suggested by previous estimates (in excess of 64 years) in order to bring our genetic estimate of long-term population size down as low as 30,000 whales (see also Roman and Palumbi 2003). If whale generation times were actually this long, it would have far reaching implications beyond the estimation of long-term population size.
Differences in time scales
Differences been genetic and catch-based estimates of past population size may arise from the fact that genetic estimates represent an average population size over evolutionary timescales, while catch-based estimates of past population size are calculated over more recent timescales. Long-term estimates of population size based on genetic data represent the weighted harmonic mean of population size over 4Ne generations (e.g., up to 4,000 generations if Ne = 1,000), but with greater weight on more recent time scales (Beerli 2009). Therefore it is possible that just prior to whaling, humpback whales were less abundant than their long-term average population size. This explanation would also need to be true of gray whales in the North Pacific (Alter and Palumbi 2009) but not minke whales in the Antarctic (Ruegg et al. 2010).
In the future, it will be important to investigate more fully how past environmental variation may have influenced long-term population size in whales and whether environmental conditions just prior to whaling would have supported a population at, above, or below the long-term average abundance. This information may be especially helpful in predicting the effect of climate change on whale populations. If, for example, whale populations were generally higher during glacial maxima and lower during glacial minimum, then as the global oceans warm and ice melts, there may be a long-term decline in whale abundance. Such long-term data would be particularly useful in assessments of current and future whale conservation status.
Comparison between catch-based and genetic-based estimates of pre-whaling abundance
There has been substantial controversy surrounding the difference between genetic and catch-based estimates of pre-whaling abundance (Lubick 2003; Holt and Mitchell 2004; Clapham et al. 2005). In order to determine whether or not inaccuracies in the catch record lead to an underestimate of the number of whales before whaling, Smith and Reeves (2010) combined previously-used sources of information with additional data from archives to fill some gaps in our understanding of North Atlantic humpback whale removals. The results of their reanalysis indicate a new overall estimate of total removals that is only 6 % higher than that used previously by the IWC Scientific Committee (30,852, SE = 655). Thus, despite a reanalysis on both sides, our multi-locus estimate of average long-term population size remains higher than the pre-whaling estimate of abundance based upon catch records.
One approach to resolving these discrepancies has been population modeling (Baker and Clapham 2004). Here, historical catches, current information about reproductive rates and modern population estimates are joined together in an analytical framework that might be able to reconcile divergent views about past populations. However, population modeling performed by Punt et al. (2006) for the North Atlantic humpback shows a poor ability to explain past population crashes and current population growth. The problem stems from the fact that North Atlantic humpback whale populations can grow so quickly [6–7 % per year (Zerbini et al. 2010), that their past populations should not have collapsed at the estimated hunting rates. For example, if North Atlantic humpback whales had an original population size of 30,000 animals, and a 6–7 % annual reproductive rate at maximum sustainable yield of 64 % of the original population size (19,200 animals), then the population as a whole should have sustained a hunt of 1,152 animals a year indefinitely. Yet, data from Smith and Reeves (2010), Fig. 1 show that there has never been a recorded catch of North Atlantic humpback whales that is this high: there were only two periods of time of a few years each when the total taken was above 400 animals per year. Even though the above estimate of sustainable yield is very crude, it demonstrates the large discrepancy between the catch record and the reproductive capacity of North Atlantic whale populations.
There are several possible explanations for these discrepancies between hunting and population growth. One alternative is that the carrying capacity of the ocean to support Atlantic humpback whales might have increased 2–3 fold during the 20th century (an assumption that has not yet been supported by data or theory), and that the maximum reproductive rate of humpback whales in the 19th century was extremely low (Punt et al. 2006). Alternatively, the models would be improved if catch rates were about twice as high as suggested by Smith and Reeves (2010). Given a higher rate of catch (about 43,000–69,000 over the course of the hunt instead of 29,000, (Punt et al. 2006, Table 1), the carrying capacity of humpback whales in the North Atlantic is estimated to be about 72,000 –117,000 (Punt et al. 2006, Table 7). In general, for the ‘alternate baseline’ scenario that Punt et al. (2006) favor, models that suggest higher original estimates fit the data better (e.g. the negative log-likelihood values (−lnL) are closer to zero, (Punt et al. 2006, Table 6). These factors suggest that a larger historical number of humpback whales in the North Atlantic would better fit the catch data, the mathematical models and the genetic data.
The summary of these various threads of evidence is that estimates of historical abundance of North Atlantic humpback whales from catch and genetic data are converging, but remain about 2–3 fold apart. Older casual catch-based estimates of original population size from before 1990 (10,000–20,000) have been superseded by population models allowing for enhanced catch rates (20,000–46,000, Punt et al. 2006). The single locus mtDNA only genetic estimate moved from 240,000 to 150,000, after correcting for mutation rate and the addition of multiple nuclear loci resulted in an estimate of ~112,000 (95 % CI 45,000–235,000). Further declines are possible if whale reproductive life times are vastly higher than currently supposed. Regardless of the generation length employed, the lower 95 % confidence limits on multi-locus estimates of long-term population size are now much closer to the range of the population models. Further closure of these differences may depend on population trajectories of whales during climate cycles, the development of population models that correctly reflect past population trajectories, and an enhanced view of whale generation times.
We thank Barry Nickel with his help with the creation of Fig. 1. This work was supported by a grant from the Lenfest Ocean Program (#2004-001492-023).