Predicting the distribution of a rare chipmunk (Neotamias quadrivittatus oscuraensis): comparing MaxEnt and occupancy models

Abstract

Species distribution models (SDMs) use presence records to determine the relationship between species occurrence and various environmental variables to create predictive maps describing the species’ distribution. The Oscura Mountains Colorado chipmunk (Neotamias quadrivittatus oscuraensis) occurs in central New Mexico and is of conservation concern due to its relict distribution and threats to habitat. We previously created an occupancy model for this taxon, but were concerned that the model may not have adequately captured the ecological factors influencing the chipmunk’s distribution because of the data hungry nature of occupancy modeling. MaxEnt is another SDM method that is particularly effective at testing large numbers of variables and handling small sample sizes. Our goal was to create a MaxEnt model for the Oscura Mountains Colorado chipmunk and to compare it with our previous occupancy model for this taxon, either to strengthen our original assessment of the relevant ecological factors or identify additional factors that were not captured by our occupancy model. We created MaxEnt models using occurrence records from baited camera traps and opportunistic surveys. We adjusted model complexity using a novel method for tuning both the regularization multiplier and feature class parameters while also performing variable selection. We compared the distribution maps and variables selected by MaxEnt to the results of our occupancy model for this taxon. The MaxEnt and occupancy models selected similar environmental variables and the overall spatial pattern of occurrence was similar for each model. Likelihood of occurrence was positively related to elevation, piñon woodland vegetation type, and topographic variables associated with escarpments. The overall similarities between the MaxEnt and occupancy models increased our confidence of the ecological factors influencing the distribution of the chipmunk. We conclude that MaxEnt offers advantages for predicting the distribution of rare species, which can help inform conservation actions.

A case for multiscale habitat selection studies of small mammals

“…Extrapolating results of a habitat selection study to a different geographic area might lead to misinterpretation of habitat information. Prior studies appropriately used occupancy modeling ( Perkins-Taylor and Frey 2018 ) and species distribution modeling ( Perkins-Taylor and Frey 2020 ) to investigate first-order habitat selection by N. q. oscuraensis ( Meyer and Thuiller 2006 ). N. q. oscuraensis selected areas at high elevation with piñon woodland and escarpments ( Perkins-Taylor and Frey 2018 , 2020 ).…”

“…Three studies evaluated microhabitat selection (i.e., third-order selection) in the eastern chipmunk ( T. striatus — Geier and Best 1980 ), the Colorado chipmunk ( N. quadrivittatus — Rivieccio et al 2003 ), and Palmer’s chipmunk ( N. palmeri ) and the Panamint chipmunk ( N. panamintinus — Lowrey and Longshore 2013 ). Three studies evaluated habitat selection at the landscape scale (i.e., first-order selection) in N. panamintinus ( Lowrey et al 2016 ) and a subspecies of the Colorado chipmunk, the Oscura Mountains Colorado chipmunk ( N. q. oscuraensis — Perkins-Taylor and Frey 2018 , 2020 ). It is possible that mammalogists have a misunderstanding about the habitat needs of many North American chipmunk species because of the lack of habitat selection studies and complete absence of multiscale habitat selection studies.…”

Linking camera‐trap data to taxonomy: Identifying photographs of morphologically similar chipmunks

“…Understudied and imperiled species are often rare, difficult to detect (Linkie et al, 2013;Thomas et al, 2020), and vulnerable to mismanagement, and so ensuring high identification accuracy for these species is of especial importance. remote-camera surveys worldwide (Wearn & Glover-Kapfer, 2019), the deployment of remote cameras in biodiversity monitoring networks that require identifications of many species (Kays et al, 2020;Steenweg et al, 2017), and the increased use of camera traps for taxonomic groups that commonly co-occur with morphologically similar species (De Bondi et al, 2010;McDonald et al, 2015;Perkins-Taylor & Frey, 2020), both the risk of misidentification and the impacts on global conservation will increase if unaddressed.…”

A Maxent Predictive Model for Hunter-Gatherer Sites in the Southern Pampas, Argentina

“…Environmental layers with collinearity will affect the interpretation of the results. To correct this parameter, some authors suggest eliminating highly correlated variables (see Perkins-Taylor & Frey 2020). Multicollinearity between predictors was investigated using ENMTools Correlation Pearson correlation coefficient (Warren, Glor & Turelli 2010).…”