Comparison of five modelling techniques to predict the spatial distribution and abundance of seabirds

By S. Oppel, Ana Meirinho, I. Ramírez, Beth Gardner, A. F. O'connell, Peter I. Miller, M. Louzao. 2012

307 citations recorded by the source corpus.

Knowledge about the spatial distribution of seabirds at sea is important for conservation. During marine conservation planning, logistical constraints preclude seabird surveys covering the complete area of interest and spatial distribution of seabirds is frequently inferred from predictive statistical models. Increasingly complex models are available to relate the distribution and abundance of pelagic seabirds to environmental variables, but a comparison of their usefulness for delineating protected areas for seabirds is lacking. Here we compare the performance of five modelling techniques (generalised linear models, generalised additive models, Random Forest, boosted regression trees, and maximum entropy) to predict the distribution of Balearic Shearwaters (Puffinus mauretanicus) along the coast of the western Iberian Peninsula. We used ship transect data from 2004 to 2009 and 13 environmental variables to predict occurrence and density, and evaluated predictive performance of all models using spatially segregated test data. Predicted distribution varied among the different models, although predictive performance varied little. An ensemble prediction that combined results from all five techniques was robust and confirmed the existence of marine important bird areas for Balearic Shearwaters in Portugal and Spain. Our predictions suggested additional areas that would be of high priority for conservation and could be proposed as protected areas. Abundance data were extremely difficult to predict, and none of five modelling techniques provided a reliable prediction of spatial patterns. We advocate the use of ensemble modelling that combines the output of several methods to predict the spatial distribution of seabirds, and use these predictions to target separate surveys assessing the abundance of seabirds in areas of regular use.

Subfields: Ecological Modeling, Ecology

View Paper: https://www.semanticscholar.org/paper/e1e2472b03680f5f821a278ff2854d7552a82957

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecological Modeling

In Ecology

Citations in the corpus

Dynamic ensemble models to predict distributions and anthropogenic risk exposure for highly mobile species (2019, methodology): “Predictions generated from multi‐model ensembles can also be more accurate than those of single models (Marmion, Parviainen, Luoto, Heikkinen, & Thuiller, 2009; Oppel et al., 2012; Scales et al., 2015).”
Identifying predictable foraging habitats for a wide‐ranging marine predator using ensemble ecological niche models (2015, background): “Nevertheless, multi-model ensembles can increase confidence in model predictions, where they are implemented with awareness of technical limitations (Marmion et al., 2009; Oppel et al., 2012).”
Predicting cetacean abundance and distribution in a changing climate (2018, unknown): “…from human activities (Benson et al., 2011; Forney, Becker, Foley, Barlow, & Oleson, 2015; Gilles et al., 2016; Hammond et al., 2013; Leathwick, Elith, Francis, Hastie, & Taylor, 2006; Louzao et al., 2006; Oppel et al., 2012; Redfern et al., 2013; Redfern, Hatch, et al., 2017; Torres et al., 2015).”
A comparison between Ensemble and MaxEnt species distribution modelling approaches for conservation: A case study with Egyptian medicinal plants (2020, background): “Ensemble modelling has been shown to improve model predictions (Grenouillet et al., 2011; Oppel et al., 2012; Stohlgren et al., 2010), reduce overfitting when modelling rare species (Breiner et al., 2016), and has been advocated as a better alternative to single models for future climate projection…”
Migratory marine species: their status, threats and conservation management needs (2014, background): “…data on species distribution, abundance and movement from at-sea surveys and satellite tracking devices with remotely sensed oceanographic environmental data are now becoming common place (Shillinger et al., 2008; Block et al., 2011; Louzao et al., 2011; Oppel et al., 2012; Mannocci et al., 2013).”
Spatial models for distance sampling data: recent developments and future directions (2013, methodology): “We note that there are many approaches to modelling spatially referenced count data (Oppel et al. 2012, provides an overview of such methods for marine bird modelling).”