Using Spatio-Temporal Species Distribution Models to Explore the Effects of Reef-Forming Coral on Seamount-Associated Fishes in New Zealand

Species distribution models (SDMs) are important tools for exploring the complex associations between species and their habitats to support marine conservation and management. In New Zealand, orange roughy and black oreo fisheries are characterised in part by their association with seamounts. New Zealand seamounts are also home to reef-forming corals. However, direct linkages between seamount-associated fishes and reef-forming corals are uncertain; that is, the extent to which coral might provide biogenic structures for shelter, feeding, spawning, or recruitment in New Zealand waters. This research aims to investigate the degree of association between two seamount-associated fish species, orange roughy (Hoplostethus atlanticus) and black oreo (Allocyttus niger), and reef-forming corals, using vector autoregressive spatio-temporal (VAST) models fitted to available monitoring data. To achieve this, we characterised the available data and single-species VAST models of data collected by observers onboard commercial bottom trawlers were selected. These models included the probability of encountering reef-forming coral as an environmental covariate. To allow for the inclusion of this environmental covariate, other (binomial) VAST models were developed and cross-validated for reef-forming coral, with alternative methods of addressing spatial bias in the coral data investigated as a secondary objective. Four models (i.e., four fish cases) were developed for orange roughy and black oreo during their spawning and non-spawning seasons. The VAST models for fishes were two-stage delta models that combined spatio-temporal estimations of probability of encounter and positive catch rate to predict fish spatio-temporal density patterns in New Zealand waters. These models accounted for unmeasured (latent) spatial and spatio-temporal variation, random vessel effects, and known environmental covariates. The effect of reef-forming coral on fishes was assessed by: (1) comparing the Akaike information criterion (AIC) of VAST models that did or did not include reef-forming coral as a covariate; (2) computing the percent deviance explained by reef-forming coral; (3) investigating the statistical significance of the reef-forming coral effect on fish probability of encounter and positive catch rate; and (4) analysing the marginal effect plots for the reef-forming coral covariate. Results indicated that AIC decreased for spawning orange roughy and non-spawning black oreo, which suggests that coral improved model parsimony in these two cases. However, the percent deviance explained by reef-forming coral was very small for the four cases considered. Across the four fish cases, reef-forming coral had a consistently significant effect on fish positive catch rates but not on fish probabilities of encounter. Ultimately, linkages between reef-forming coral and orange roughy and black oreo during their spawning and non-spawning seasons remain quite unclear as marginal effects plots often suggested no meaningful relationship and were associated with considerable uncertainty. This is not to say, however, that this dynamic does not exist, but rather that more targeted work is required in New Zealand to understand interactions between key fish species and their habitat.