Current State of the Art for Statistical Modelling of Species Distributions

Predictions of 27 Arctic pelagic seabird distributions using public environmental variables, assessed with colony data: a first digital IPY and GBIF open access synthesis platform

“…Major impacts are introduced through human provisioning of additional food supplies (Furness and Monaghan 1987), but these effects are poorly studied. Further, virtually none of the existing seabird distribution maps from scientific sources are in entire agreement with each other, the underlying raw data are rarely available, and the derived maps are not quantitative, nor based on consistent data and latest techniques and accuracy assessments that are usually used in the disciplines of species distribution and predictive modeling (e.g., Guisan and Zimmermann 2000;Pearce and Ferrier 2000;Hegel et al 2010). Statistically, the literature sources rarely perform well when scrutinized, yet this is required for a sound science-based management of a precious natural resource (Cushman and Huettmann 2010).…”

Comparing species distribution models: a case study of four deep sea urchin species

“…New management approaches such as Marine Protected Areas (MPAs) and new environmental legislation such as the Marine Strategy Framework Directive (MSFD; 2008/56/EC) require information about the distribution of species and their habitat (Borja et al, 2010;Sundblad et al, 2011). In this context, species distribution models (SDMs), also called habitat suitability models, constitute a very useful tool to obtain habitat maps based on species distribution records and environmental layers of geographic information (Fielding & Bell, 1997;Bryan & Metaxas, 2007;Hegel et al, 2010;Reiss et al, 2011). These models are based on the concept of the ecological niche, defined as the set of environmental conditions within which a species can survive and persist (Hutchinson, 1957).…”

“…Although density/abundance records have been used in SDMs (e.g., Oppel et al, 2012;Darr et al, 2014), in most of the cases these models analyze the distribution of a species using presence/absence or presence-only data (Hegel et al, 2010). Comparing both approaches in marine ecosystems is relevant since the use of presence-only models in the marine environment has increased markedly during the last few years (e.g., Bryan & Metaxas, 2007;Galparsoro et al 2009;Monk et al, 2010;Hermosilla et al, 2011;Reiss et al, 2011;Pierrat et al, 2012;Ross & Howell, 2013;Martín-García et al, 2013;García-Alegre et al, in press) although the reasons behind choosing this type of method and the consequences of not including the absences are not always clear (Dorazio, 2012;Monk, 2013;Yackulic et al, 2013).…”

Area utilization patterns of humpback dolphins (Sousa plumbea) in Richards Bay, KwaZulu-Natal, South Africa

“…Subsampling of datasets is often used to reduce autocorrelation, although De Solla et al (1999) finds this to be less effective than alternative approaches, and Cushman et al (2005) and indicate that trying to attain statistical independence through subsampling ''incurs heavy costs in terms of information loss''. Mixed effect models are said to be a better choice when dealing with lack of independence in the data (Millar and Anderson 2004;Dormann et al 2007;Chaves 2010;Hegel et al 2010), and the random factor structure accounts better for autocorrelated error variances (Økland 2007) and reduces overall variance (Hegel et al 2010). Random factors deals better with unbalanced sample designs, such as this study (Hegel et al 2010), and it has been frequently emphasised that one should only conclude general interpretations from the predictions derived from these mixed effect models (Millar and Anderson 2004;Venables and Dichmont 2004;Økland 2007;Chaves 2010).…”