Our results show that GoM BFT larval survey samples can provide the crucial mark events for eventual CKMR estimates of adult abundance. The adult parents marked by larval samples can be directly recaptured in the fishery many years later as POPs, and indirectly through their progeny in future samples of larvae, as evidenced by the two cross-cohort HSPs (XHSPs) recovered in this study, which imply that a parent survived and spawned in the GoM in consecutive years. As more cohorts are sampled in future, the growing number of XHSPs could be used to estimate average adult survival rates, in addition to helping with the estimation of adult abundance31, as is now done for southern blue tuna40.
There is a modest level of sibship within our 2016 samples, and a high level (involving over half the samples) in 2017, but it turns out not to be high enough to cause serious problems for POP-based CKMR. High sibship per se does not lead to bias in CKMR by virtue of the statistical construction of the estimate, but it does increase variance, which can be summarized through a reduction in effective sample size. In a POP-based CKMR model, our effective sample size would be about 75% of nominal for the two years combined, or 66% of nominal for the targeted sampling of 2017. Since it is actually the product of adult and juvenile sample sizes which drives precision in CKMR14, one way to think about the 75% is that we will need about 33% more adult samples to achieve a given precision on abundance estimates than if we had somehow been able to collect the same number of “independent” juvenile samples (i.e. without oversampling siblings). That increase is appreciable but entirely achievable; for WBFT, it is logistically much easier to collect more feeding-ground adult samples than to collect more larvae, and at present there is no known practical way to collect large numbers of older, more dispersed, and thus more independent, juvenile western origin bluefin tuna (WBFT).
This study was motivated by the concern that sibship might be a serious impediment to use of WBFT larvae for CKMR. High levels of sibship have been found in larval collections for other taxa despite a pelagic larval phase, suggesting that abiotic factors can impede random mixing of larvae after a spawning event41. Our larval samples were only a few days old (4–11) and thus had little time to disperse since fertilization; our concern beforehand was that each tow might sample the offspring of a very small number of adults (one spawning group in one night), and in 2017 that repeatedly towing the same water mass might simply be resampling the same “family”. In practice, though, the cumulative effect was limited. Samples were not dominated by progeny from just a few adults; the maximum DPG size (i.e., number of offspring from any one adult) was 5, which is under 2% of the larval sample size. There are several possible reasons for this finding. First, plankton sample tows are typically standardized to a ten-minute duration, covering on average about 0.3 nautical miles. Based on continuous plankton cameras42, each tow is likely to tow through multiple patches of zooplankton, and therefore potentially multiple patches of BFT larvae. Second, spawning aggregations of BFT may contain many adults. For example, on the spawning grounds near the Balearic Islands in the Mediterranean, purse seine fisheries target spawning fish and individual net sets routinely capture upwards of 500 mature individuals43. These numbers suggest that BFT spawner aggregations can be quite large, although the number of individuals that contribute gametes to a single spawning event may be lower. The results of this study pose intriguing scenarios for understanding BFT larval ecology and spawning behavior, which could be explored with larger sample sizes paired with data on oceanographic conditions, direct observation of spawning aggregations, and modeling to compare observed and predicted dispersal. The results of this study are based on just two years of sampling, and numerous practical and theoretical challenges remain to fully understand BFT reproduction in the GoM.
Our sibship impact calculations assume use of an unmodified adult-size-based CKMR POP model, where each juvenile is compared to each adult taking into account the latter’s size (e.g.,14). That will give unbiased estimates, which we regard as essential in a CKMR model. However, for WBFT the estimates are not fully statistically efficient, in that some adults receive more statistical weight than others because they are marked more often (by having a large DPG), and thus variance might not be the lowest achievable. Modifying the model to fix that would be simple in a “cartoon” CKMR setting where all adults are identical (e.g., Fig. 1 of14), simply by first condensing each DPG to a single representative, then only using those representatives (rather than all the larvae) in POP comparisons. Each marked parent then receives the same weight, giving maximum efficiency. For the cartoon, this condensed-DPG model still gives an unbiased estimate of abundance, because each DPG has one parent of given sex, and the chance of any sampled cartoon adult of that sex being that parent is 1/N. The DPG-condensed effective sample size is simply half the number of distinct parents, which would be a little larger than the effective sample sizes for the unmodified model shown in Table 3; e.g., in 2017, 504/2 = 252 versus 209. However, no such straightforward improvement is available for an adult-size-based CKMR model such as is needed for WABFT. Using condensed DPGs directly would bias the juvenile sampling against larger more-fecund adults, whose DPGs will tend on average to be larger and thus to experience disproportionate condensation. Those adults would be marked less often by the DPG-condensed juveniles than the model assumes, violating the basic requirements for unbiased CKMR in14. A more sophisticated model might be able to combine unbiasedness with higher efficiency but, since the unmodified adult-size-based POP model that we expect to use is unbiased and only mildly inefficient (at worst 209/252 = 83% efficient, in 2017) there seems no particular need for extra complications at present. However, that may not hold true if we eventually move to a POP + XHSP model, where the impact on unmodified CKMR variance is worse (though there is still no bias, for the same reason as with POPs). Intuitively, the biggest impact that a DPG of size 5 can have in a POP model is to suddenly raise the number of POPs by 5 if its parent happens to be sampled; within a useful total of, say, 75 POPs, the influence is not that large. But if two DPGs both of size 5 in different cohorts happen to share a parent, then the total of XHSPs suddenly jumps by 25— likely a substantial proportion of total XHSPs. Supplementary Material B also includes effective sample size formulae for a simplified XHSP-only model, which demonstrate the increased impact of within-cohort sibship; for our WBFT samples, it turns out that the XHSP-effective size is slightly lower for the targeted 2017 samples (110) than for the 2016 samples (130), unlike the POP-only effective size. Dropping from a maximum theoretical effective sample size of 252 (half the number of DPGs) down to 110 would be rather inefficient and would increase the number of years of sampling required to yield a useful XHSP dataset. This motivates developing a modified POP + XHSP model that retains unbiasedness without sacrificing too much efficiency. In principle, that can be done by condensing each DPG but then conditioning its comparison probabilities on the DPG’s original size, in accordance with the framework in14. This is a topic for subsequent research, and the results will inform future sampling strategy decisions for WBFT.
One potential difficulty for western BFT CKMR might occur if a substantial proportion of animals reaching maturity are the offspring of “Western” (in genetic terms) adults who persistently spawn in the western North Atlantic but outside the GoM. However, as long as the adults marked by GoM larvae are well mixed at the time of sampling with any western adults that do spawn outside of the GoM, the total POP-based population estimate of genetically-western BFT from CKMR will remain unbiased. Given evidence from tagging of widespread adult movements within the western North Atlantic2, good mixing in the sampled feeding grounds seems likely; so, even if successful non-GoM western BFT spawning really is commonplace, there should not be a problem with relying on GoM larvae for at least the POP component of CKMR14.
Studies of fish early life history have long been considered to have great potential to provide novel insight into the unique population dynamics of fishes44,45,46. Sampling efforts aimed at estimating fish recruitment dynamics have spawned a diversity of larval survey programs. Examples of these long-term programs include the California Cooperative Oceanic Fisheries Investigations, International Council for the Exploration of the Sea (ICES) surveys in the North Atlantic and adjacent areas, Southeast Monitoring and Assessment Program (SEAMAP) in the GoM, Ecosystem Monitoring (EcoMon) in the Northeast U.S., and numerous others, many of which provide indices of larval abundance widely used in fisheries and ecosystem assessments. Yet, as a result of the inherent patchiness of larvae42, sampling variability, and highly variable density dependent mortality45, fisheries scientists have often struggled to determine how larval surveys relate to the adult fish populations. Inclusion of estimates of sibship among larvae collected in surveys could refine estimates of adult spawning stock biomass estimated from these surveys.
The results of this study also represent products of decades of work and coordination in obtaining high-quality DNA from larval specimens. Key steps to successful genotyping of larvae include ensuring that larvae are preserved, sorted, and handled in 95% non-denatured ethanol. In addition, strict instrument cleaning protocols must be followed, and stomachs should be removed or avoided (this study used larval tails and, when possible, eyes to avoid cross contamination of prey contents, including possible congeners and other BFT individuals). Exposure to hot lamps during the sorting and dissection processes should also be minimized to ensure that DNA quality is sufficiently high for genotyping-by-sequencing. Although the tissues available for genetic analysis were limited by the needs of other experiments that required BFT tissues, otoliths, gut contents, and other information from the same larvae, we were able to successfully genotype most larvae greater than 6 mm SL and identify thousands of informative SNPs. The lower size limit of larvae could likely be decreased if whole specimens were available for genotyping, although the use of younger larvae could increase the incidence of sibship.
In summary, while we observed both FSPs and HSPs in larval collections, with elevated sibship overall and with siblings being more prevalent within tows and in nearby tows, the level of sibship was sufficiently low that collections of GoM BFT larvae can still provide the critical genetic mark of parental genotypes required for CKMR. Our results demonstrate a crucial proof of concept and are the first step towards an operational CKMR modelling estimate of spawning stock abundance for western BFT.
Source: Ecology - nature.com
