@article{fdi:010082166,
  title = {{C}ombining regional habitat selection models for large-scale prediction : circumpolar habitat selection of {S}outhern {O}cean humpback whales},
  author = {{R}eisinger, {R}. {R}. and {F}riedlaender, {A}. {S}. and {Z}erbini, {A}. {N}. and {P}alacios, {D}. {M}. and {A}ndrews-{G}off, {V}. and {D}alla {R}osa, {L}. and {D}ouble, {M}. and {F}indlay, {K}. and {G}arrigue, {C}laire and {H}ow, {J}. and {J}enner, {C}. and {J}enner, {M}. {N}. and {M}ate, {B}. and {R}osenbaum, {H}. {C}. and {S}eakamela, {S}. {M}. and {C}onstantine, {R}.},
  editor = {},
  language = {{ENG}},
  abstract = {{M}achine learning algorithms are often used to model and predict animal habitat selection-the relationships between animal occurrences and habitat characteristics. {F}or broadly distributed species, habitat selection often varies among populations and regions; thus, it would seem preferable to fit region- or population-specific models of habitat selection for more accurate inference and prediction, rather than fitting large-scale models using pooled data. {H}owever, where the aim is to make range-wide predictions, including areas for which there are no existing data or models of habitat selection, how can regional models best be combined? {W}e propose that ensemble approaches commonly used to combine different algorithms for a single region can be reframed, treating regional habitat selection models as the candidate models. {B}y doing so, we can incorporate regional variation when fitting predictive models of animal habitat selection across large ranges. {W}e test this approach using satellite telemetry data from 168 humpback whales across five geographic regions in the {S}outhern {O}cean. {U}sing random forests, we fitted a large-scale model relating humpback whale locations, versus background locations, to 10 environmental covariates, and made a circumpolar prediction of humpback whale habitat selection. {W}e also fitted five regional models, the predictions of which we used as input features for four ensemble approaches: an unweighted ensemble, an ensemble weighted by environmental similarity in each cell, stacked generalization, and a hybrid approach wherein the environmental covariates and regional predictions were used as input features in a new model. {W}e tested the predictive performance of these approaches on an independent validation dataset of humpback whale sightings and whaling catches. {T}hese multiregional ensemble approaches resulted in models with higher predictive performance than the circumpolar naive model. {T}hese approaches can be used to incorporate regional variation in animal habitat selection when fitting range-wide predictive models using machine learning algorithms. {T}his can yield more accurate predictions across regions or populations of animals that may show variation in habitat selection.},
  keywords = {ensembles ; habitat selection ; machine learning ; prediction ; resource ; selection functions ; telemetry ; humpback whale ; {M}egaptera novaeangliae ; {OCEAN} {AUSTRAL}},
  booktitle = {},
  journal = {{R}emote {S}ensing},
  volume = {13},
  numero = {11},
  pages = {2074 [23 p.]},
  year = {2021},
  DOI = {10.3390/rs13112074},
  URL = {https://www.documentation.ird.fr/hor/fdi:010082166},
}