Data stochasticity and model parametrisation impact the performance of species distribution models: insights from a simulation study DOI Creative Commons
Charlotte Lambert, Auriane Virgili

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 19, 2023

Abstract Species distribution models (SDM) are widely used to describe and explain how species relate their environment, predict spatial distributions. As such, they the cornerstone of most planning efforts worldwide. SDM can be implemented with wide array data types (presence-only, presence-absence, count…), which either point- or areal-based, use a environmental conditions as predictor variables. The choice sampling type well resolution recognized crucial importance, yet we lack any quantification effects these decisions may have on reliability. In present work, fill this gap an unprecedented simulation procedure. We simulated 100 possible distributions two different virtual in regions. were modelled using segment- areal-based five resolutions conditions. performances inspected by statistical metrics, model composition, shapes relationships prediction quality. provided clear evidence stochasticity modelling process (particularly relationships): dataset from same survey, region could yield results. Sampling had stronger than final relevance. effect coarsening was directly related resistance features changes scale: failed adequately identify when targeted diluted coarsening. These results important implications for community, backing up some commonly accepted choices, but also highlighting up-to-now unexpected (stochasticity). whole, work calls carefully weighted implementing models, caution interpreting Figure

Language: Английский

Pharmaceutical residues in stranded dolphins in the Bay of Biscay DOI
Maitane Alzola-Andrés, Daniel Červený, Saioa Domingo‐Echaburu

et al.

The Science of The Total Environment, Journal Year: 2023, Volume and Issue: 912, P. 168570 - 168570

Published: Nov. 17, 2023

Language: Английский

Citations

3
Delayed response to environmental conditions and infra-seasonal dynamics of the short-beaked common dolphin distribution DOI Creative Commons
Charlotte Lambert, Matthieu Authier, Arny L. Blanchard

et al.

Royal Society Open Science, Journal Year: 2022, Volume and Issue: 9(11)

Published: Nov. 1, 2022

Cetaceans adjust their distribution and abundance to encountered conditions across years seasons, but we poorly understand such small-scale changes for many species, especially in winter. Crucial challenges confront some populations during this season, as the high levels of fisheries-induced mortality faced by common dolphin ( Delphinus delphis ) Northeast Atlantic shelves. For understanding winter fine-scale dynamics is crucial. We aimed identify drivers winters 2020 2021, with a focus on determining lag between oceanographic distribution. The were related temporal delays specific nature cascading effects that processes had trophic chain. By most important lags distributions, shed light understood intrusions dolphins within coastal waters winter: they displayed strong preference coastal-shelf front extensively followed its spatial variations, overall densities increasing over period peaking March–April. results presented here provide invaluable information should inform management decisions help reduce unsustainable mortalities species by-catch fisheries.

Language: Английский

Citations

5
Data stochasticity and model parametrisation impact the performance of species distribution models: insights from a simulation study DOI Creative Commons
Charlotte Lambert, Auriane Virgili

Peer Community Journal, Journal Year: 2023, Volume and Issue: 3

Published: April 7, 2023

Species distribution models (SDM) are widely used to describe and explain how species relate their environment predict spatial distributions. As such, they the cornerstone of most planning efforts worldwide. SDM can be implemented with a wide array data types (presence-only, presence-absence, count...), which either point- or areal-based, use environmental conditions as predictor variables. The choice sampling type well resolution recognized crucial importance, yet we lack any quantification effects these decisions may have on reliability. In present work, fill this gap an unprecedented simulation procedure. We simulated 100 possible distributions two different virtual in regions. were modelled using segment- areal-based five resolutions conditions. performances inspected by statistical metrics, model composition, shapes relationships prediction quality. provided clear evidence stochasticity modelling process (particularly relationships): dataset from same survey, region could yield results. Sampling had stronger than final relevance. effect coarsening was directly related resistance features changes scale: failed adequately identify when targeted diluted coarsening. These results important implications for community, backing up some commonly accepted choices, but also highlighting up-to-now unexpected (stochasticity). whole, work calls carefully weighted implementing models, caution interpreting

Language: Английский

Citations

1
Exploring environmental and biological drivers of cetacean occurrence in the cross-border region of the Malin Shelf using data from a European fishery survey DOI Creative Commons
Morgane Pommier,

Ciarán O’Donnell,

Cynthia Barile

et al.

Frontiers in Marine Science, Journal Year: 2023, Volume and Issue: 10

Published: Nov. 15, 2023

Irish and Scottish waters are important habitats for cetaceans in Europe. Yet, little data is available the region of Malin Shelf, north Ireland. Despite a rich species diversity, relative cetacean abundance appears low compared to hotspots documented west Scotland Whether this perceived prevalence accurately portrays an ecological discontinuity or arises from lack published results survey effort that transborder area remains unclear. Here, we used sighting records multi-disciplinary fisheries survey, Western European Shelf Pelagic Acoustic Survey (WESPAS), explore habitat preferences over Hebridean shelves. Northern minke whale common dolphin occurrence was modelled within Bayesian Additive Regression Trees (BART) framework, against selected environmental biological variables. No correlation observed between presence in-situ prey biomass. Minke whales distribution better explained by oceanography, notably proxies frontal activity, primary productivity. Common dolphins similarly showed shelf 5-25km fronts, but also affinities fine substrates. Favourable identified models were consistent with literature around Hebrides shed light on potentially areas along Islay front Donegal, so far unreported due deficiency. Results will contribute towards informing future monitoring, strategic management conservation efforts cross-border region.

Language: Английский

Citations

1
Data stochasticity and model parametrisation impact the performance of species distribution models: insights from a simulation study DOI Creative Commons
Charlotte Lambert, Auriane Virgili

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Jan. 19, 2023

Abstract Species distribution models (SDM) are widely used to describe and explain how species relate their environment, predict spatial distributions. As such, they the cornerstone of most planning efforts worldwide. SDM can be implemented with wide array data types (presence-only, presence-absence, count…), which either point- or areal-based, use a environmental conditions as predictor variables. The choice sampling type well resolution recognized crucial importance, yet we lack any quantification effects these decisions may have on reliability. In present work, fill this gap an unprecedented simulation procedure. We simulated 100 possible distributions two different virtual in regions. were modelled using segment- areal-based five resolutions conditions. performances inspected by statistical metrics, model composition, shapes relationships prediction quality. provided clear evidence stochasticity modelling process (particularly relationships): dataset from same survey, region could yield results. Sampling had stronger than final relevance. effect coarsening was directly related resistance features changes scale: failed adequately identify when targeted diluted coarsening. These results important implications for community, backing up some commonly accepted choices, but also highlighting up-to-now unexpected (stochasticity). whole, work calls carefully weighted implementing models, caution interpreting Figure

Language: Английский

Citations

0