Ground Validation Reveals Limited Applicability of Species Distribution Models for Dakota Skipper (Hesperia dacotae, Lepidoptera: Hesperiidae) Recovery Efforts in Canada

Abstract

Species distribution models (SDMs) use spatial relationships between species occurrence and habitat (predictor) variables to generate maps of habitat suitability across a region of interest. These maps are frequently used in recovery planning efforts for endangered species, but they are influenced by data availability, selection of predictor variables, and choice of model type. Ground validation is necessary to robustly evaluate map accuracy, but it is rarely done, making it difficult to determine which modeling approach is best-suited for a given species or region. To address this uncertainty, we used two SDM types (Maxent and GLM) and two methods of selecting predictor variables to build four SDMs for an endangered prairie butterfly (Dakota skipper, Hesperia dacotae) in two regions of Manitoba, Canada. We then conducted field-based habitat suitability assessments at 120 locations in each region to enable direct comparisons of model output and accuracy. We found that soil type and surrounding landcover (grassland versus cropland) were important predictors of species occurrence regardless of region, predictor selection method, or model type. Cross-validation statistics indicated that most SDMs performed well (AUC > 0.7), but ground validation revealed that the habitat suitability maps they generated were inaccurate (Cohen’s kappa < 0.4). Maxent models produced more accurate maps than GLMs, likely because false species absences adversely affected the latter, but only one Maxent-based map was accurate enough to help locate sites for future field investigations (Cohen’s kappa > 0.3). Our results emphasize the importance of ground-validating SDM-based habitat suitability maps before incorporating them into species recovery plans.

Comparing Two Species Distribution Models Using Satellite Only and Ready-Made Environmental Variables for the Dakota Skipper (Hesperia dacotae), Interlake Region of Manitoba, Canada

“…Larger values are also selected to increase confidence that suitable areas are in fact suitable at the expense of excluding some areas (Liu et al 2005). Model output results have also been used to apply a more objective threshold with some degree of success (Jiménez-Valverde and Lobo 2007), however some question the utility of thresholding and instead suggest using tests which incorporates the continuous nature of the relative likelihood surface (Merow et al 2013;Dearborn et al 2022).…”

“…Field validation involves gathering a test data set by surveying sites within the study area and assigning a species presence likelihood or suitability score to the sites based on site attributes which are correlated to species presence. Similar to model statistic evaluation, the test data is then compared to the relative likelihood values from the response surface by applying thresholds and also by using the continuous relative likelihood values (Dearborn et al 2022).…”

“…As part of this study, I also validated the models by ground truthing model results, which has been shown to be a better indicator of model performance than by evaluating only model output statistics (Westwood et al 2019;Dearborn et al 2022) and which the other studies did not perform. The objectives of this project were to: 1) Compare Dakota skipper models using freely available high resolution remotely sensed products to those using more traditional environmental predictors.…”

“…A common approach to obtain these model variables is to use "pre-made" readily available physical/environmental spatial products the user is familiar with (e.g. Land use/land cover, soil, Forest Resource Inventory, digital elevation model, and climatic variables) (Rabe et al 2002;Wintle et al 2005;Royer et al 2008;Elith and Leathwick 2009;Lahoz-Monfort et al 2010;Lapin et al 2013;Westwood et al 2019;Dearborn et al 2022). A different, or complementary approach, which is gaining traction with the greater accessibility of no-cost satellite imagery, is to create model variables solely from satellite imagery (Lahoz-Monfort et al 2010).…”

“…Past Dakota skipper research in Manitoba, Saskatchewan and the northern U.S. has generally focused on describing local biotic and abiotic site characteristics (Dana 1991;Rigney 2013;Seidle et al 2018) as well as some preliminary SDM studies to determine potential habitat range of Dakota skipper and Poweshiek skipperling, Oarisma poweshiek (Westwood et al 2019;Post van der Burg et al 2020;Seidle et al 2020;Dearborn et al 2022) including additional modelling response to climate change (Barnes et al 2024).…”