AbstractMethods to reliably identify jump discontinuities in biological time series and to assess the specific contribution of various covariates are rapidly progressing. Here, we took advantage of these statistical improvements as well as those seen in complementary, downâscaled climate and biogeochemical models to investigate causes of the substantial interannual variability observed in recruitment strength in hindcast analyses. This systematic metaâanalysis included 23 dataârich, commercially valuable, warmâ and coldâtemperate stocks in the North, Norwegian and Barents Seas. Since this study focuses on recruitment strength variability, we have used the term ârecruitment regime shiftâ to distinguish from the concept of ecosystem regime shift. The breakpoint analysis revealed that the former criterion applied to more than half of the time series, mainly with respect to North Sea stocks but also to those in the Norwegian Sea. The exploratory analysis using vcGAM indicated that 1â3 shifts per stock were real, when using five drivers spanning spawning stock biomass to largeâscale climatic processes. Thus, nonâstationary relationships were extensively prevalent, indicating that each stock is uniquely adapted to its locally varying conditions. Outputs from the stationary GAM resembled those from the vcGAM but not after the threshold year. Inâdepth case studies showed that the proxy of a given driver for the process which was to be included should be critically considered in a spatiotemporal context. Furthermore, the stockâspecific uncertainty associated with the given recruitment figures as such should also be an inâbuilt component of the model construct and thereby in the evaluation of the output.
Read full abstract