ABSTRACT
Chapter 9 introduces the concept of ecological teleconnections, whereby individual movement or other mechanisms cause densities to by strongly correlated at geographically distant locations. It then introduces Empirical Orthogonal Functions (EOF), where spatio-temporal dynamics are summarized as the product of a spatial response map and an estimated temporal index. It then generalizes this conventional EOF using a generalized linear latent variable model (EOF-GLLVM). Using a case study involving summer sea ice concentrations in the Arctic Ocean, the chapter shows that EOF-GLLVM can accurately reconstruct aggregate spatio-temporal dynamics while also highlighting dominant modes of ecosystem variability. Finally, it introduces “confirmatory factor models” (CFA) where EOF indices are replaced with an hypothesized time-series covariate. Using this CFA, Chapter-9 highlights that the spatial extent of near-freezing water at the bottom of the eastern Bering Sea can explain multi-decadal dynamics for eastern Bering Sea pollock.
