Reply to Roberts et al.: Faith in no fishery impact on New Zealand sea lions based on misunderstandings and unsubstantiated claims

In Meyer et al. (1), we fitted a Bayesian state-space Gompertz population model to assess whether the implementation of sea lion exclusion devices (SLEDs) in trawl fisheries has served to protect New Zealand sea lions (NZSLs) or obscured ongoing mortality. We found strong evidence of obscured ongoing mortality. Roberts et al. (2) do not take issue with our main analysis but criticize two of our supplemental results. Here, we provide clarification of our work and show that Roberts et al.’s (2) claims are based on misunderstandings.

First, Roberts et al. (2) criticize our sensitivity analysis, suggesting it only used the upper 95% confidence interval bounds of interaction rates (IRs). This is incorrect. Rather, we allowed total interactions to take any possible value in the interval of reported upper and lower confidence limits. Our sensitivity analysis indicates that the uncertainty in interaction estimates, which Roberts et al. (2) highlight in their figure 1 C and D, had negligible effects on our original results [see SI Appendix, figure S8 in Meyer et al. (1)]. Additionally, Roberts et al. (2) do not provide any evidence that alternative sensitivity analyses alter our findings.

Second, Roberts et al. (2) focus on the simple linear regression of population growth rate versus IR, emphasizing a nonsignificant effect of IRs. However, they provide no information on effect size and statistical power, mandatory information to interpret a negative statistical result as evidence of no biological effect versus a weak statistical analysis. Without that information their negative result remains uninformative rather than countering evidence to our main findings.

Roberts et al. (2) reassert two key assumptions underlying the current stock assessment model for NZSL management that sea lions that escape trawls via SLEDs are likely to survive the encounter and that NZSLs experience negative density-dependent population growth. We carefully discuss these assumptions in our paper (1) and explain why we think these are flawed and lead to a predetermined result in the stock assessment model that trawl fisheries have no effect on NZSLs. Roberts et al. (2) do not offer new information or explanations to respond to these concerns.

We agree with Roberts et al. (2) that factors other than fishery mortality affect population dynamics. However, the point of our analysis was to evaluate whether SLED implementation has helped NZSL recovery versus the alternative outcome of obscured mortality being correlated with NZSL population decline. That there are also other factors influencing NZSLs does not undermine this result unless those other factors are highly correlated with the IR, but there is no evidence that this is the case.

We are surprised by Roberts et al.’s (2) assertion that the NZSL decline has ceased in recent years. Clearly, this is at odds with the recent uplisting of the NZSL to an endangered status (3). Over the study period, the declining NZSL population has experienced periods of stable or even increasing pup production, which, as we show in our original paper (1), corresponds to periods of low fisheries impact.