Predicting the distribution of non‐vagile taxa: a response to McNyset and Blackburn (2006) and re‐evaluation of Stockman et al. (2006)

Abstract

This paper addresses the issues raised by McNyset and Blackburn (2006) in their response to Stockman et al . (2006). Re-evaluation of our published GARP analyses by McNyset and Blackburn showed that a much improved ecological niche model is obtained for predicting the distribution of the trapdoor spider genus Promyrmekiaphila in central/northern California. The improved niche model results in a substantially reduced omission error rate and a predictive model comparable to models obtained using other methods (GLM and BIOCLIM). However, the improved GARP models have a high commission error rate (> 0.75); consequently, the inferences regarding difficulties in modelling non-vagile taxa drawn by Stockman et al . remain valid. Finally, we discuss other relatively minor criticisms of our study raised by McNyset and Blackburn and issues related to the peer review of our original paper. KeywordsAraneae, Desktop GARP, niche modelling, predicting species occurrences, Promyrmekiaphila . McNyset and Blackburn (2006) reassessed the data and analyses reported in our paper (Stockman et al ., 2006). Their use of Desktop GARP (DG, Scachetti-Pereira, 2002) with our data produces a more accurate niche model, particularly with respect to omission rate, than the one we obtained. The primary criticism raised by McNyset and Blackburn (2006) resides with our mistaken use of many single-specimen models summed to form a single predictive model. Secondary criticisms include use of categorical layer data (soil and vegetation class), model comparison, use of layer data at too fine a resolution, and use of multiple species to generate single models. The improvements found in their DG models are largely attributed to the use of all 42 of our data points to generate a summed predictive model. While we do not necessarily agree with all of the points raised by McNyset and Blackburn we are concerned with the discrepancy between our original analysis and the one they provide. We report here a reanalysis of our data set (Stockman et al ., 2006) for DG, address the later concerns raised by McNyset and Blackburn, discuss ground-truthing of our models, and make some final concluding remarks. This paper is intended to serve as both a response to McNyset and Blackburn (2006) and as a correction to Stockman et al . (2006). Our original conclusion that non-vagile organisms present a special class of problems for predictive niche modelling is not only retained in light of the corrected model, but is further strengthened.We would like to state upfront explicitly that the DG predictive models we report here are consistent with those of McNyset and Blackburn. We trace the discrepancy to a data input mistake, rather than 'ignorance' , at one of the earlier stages in our analysis. We were likewise able to generate DG models for Promyrmekiaphila distributions that were much improved with respect to omission rate. The intent of our original paper was not to evaluate the DG software package nor was it intended as an evaluation of the GARP algorithm; in fac...

Geographic distribution of Africanized honeybees (Apis mellifera) reflects niche characteristics of ancestral African subspecies

“…In this study, the Genetic Algorithm for Rule Set Production (GARP), an algorithm that has shown good predictive ability of the potential distributions of species (Stockwell & Peterson 2002;Anderson et al 2003) is applied. Whilst GARP performance seems to be inferior when compared with new methods (e.g., MAXENT) under the "typical" circumstance of modeling a species' distribution on its native habitat (Elith et al 2006; also, see Stockman et al (2006) and references therein, for a discussion over an evaluation of GARP models), its ability to predict a distribution in unsampled regions (known as transferability) (Peterson et al 2007) make it a suitable choice for this study. GARP has been widely applied to help understand patterns of species invasions (Peterson & Vieglais 2001;Peterson 2003;Peterson & Robins 2003), both to predict future invasions or with the purpose of understanding past processes.…”

Areas of endemism of Mexican mammals: reanalysis applying the optimality criterion

“…Many algorithms to model ecological niches have been devised (Guisan & Zimmermann, 2000;Guisan & Thuiller, 2005) and some papers comparing their performance have been published (Elith et al, 2005;Hernández et al, 2006;Stockman, Beamer & Bond, 2006;Sergio et al, 2007;Tsoar et al, 2007). Although there is no agreement over the superiority of an algorithm over others when modelling ecological niches, the genetic algorithm for rule-set prediction (GARP) has been predominantly used in parsimony analyses of endemicity and has proven to be appropriate for predicting mammal species distributions (Illoldi-Rangel, Sánchez-Cordero & Peterson, 2004;Sánchez-Cordero et al, 2005a, b).…”

Can species distribution modelling provide estimates of population densities? A case study with jaguars in the Neotropics

“…, 2009; Roura‐Pascual et al. , 2009) or compare the results, testing their performance to evaluate which one provides the best estimation for the species and for the research purposes (McNyset & Blackburn, 2006; Stockman et al. , 2006a,b; Bedia et al.…”