Ecosphere, Journal Year: 2024, Volume and Issue: 15(4)

Published: April 1, 2024

Abstract Numerous spatiotemporal species distribution modeling frameworks are now available to the ecological practitioner. This study compared three such accessible in R programming language: generalized additive models with smooths as implemented by mgcv, linear mixed based on nearest neighbor Gaussian processes starve, and stochastic partial differential equations approach sdmTMB. The primary focus was compare inferences obtained from applying these case of orange‐footed sea cucumber, Cucumaria frondosa , Scotian Shelf off Nova Scotia, Canada. Each model fit catch data (2000–2019) Fisheries Oceans Canada's annual Research Vessel Snow Crab surveys. Environmental covariates were sourced high‐resolution layers, including physical oceanographic, bathymetric, seafloor morphometric datasets. captured variability cucumber that would have been overlooked without a approach. Although their predictions similar, within C. spatial reserves, provided different regarding covariate effects. suggests while practitioners primarily interested mapping distributions need only apply most familiar framework, those concerned identifying predictive environmental may benefit comparing output multiple approaches. Employing approaches can also serve validation technique.

Language: Английский