Reconstruction of the historical range alters niche estimates in an endangered rodent

Abstract

Defining historical baselines is critical for species conservation. Under the niche reduction hypothesis, species in decline may be restricted disproportionately from parts of their environmental niche. This bias likely has important implications for modeling species’ distributions if only contemporary occurrences (i.e. post‐range reduction) are used, because suitable habitat will be classified as unsuitable. Unfortunately, robust historical occurrence data is rarely available for sensitive species. In this study, we documented historical locations of the endangered, keystone giant kangaroo rat Dipodomys ingens by examining aerial imagery for burrow mounds. These burrow mounds are readily identifiable and distinguishable from other soil disturbances. We found giant kangaroo rat burrows well outside the currently accepted estimate of their historical distribution. Following the niche reduction hypothesis, we found that giant kangaroo rats have been extirpated from the flattest, hottest, driest parts of their range due to agricultural conversion. This reduction in their realized niche led to significant changes between historical and contemporary models of their distribution. We found that giant kangaroo rats may have occupied up to 56% more habitat historically than currently believed. Our results provide new guidance for managers working on restoration and habitat protection for this ecosystem engineer. This study highlights the critical importance of modeling historical distributions using the entire environmental niche once occupied by species of conservation need.

Combining Occurrence and Habitat Suitability Data Improve Conservation Guidance for the Giant Kangaroo Rat

“…Second, we visually assessed aerial and satellite imagery for evidence of burrow mounds made by giant kangaroo rats. Third, we obtained a model of habitat suitability for giant kangaroo rats trained with occurrence data from before their wide‐spread decline (Rutrough et al 2019) to identify areas of agreement and disagreement with these estimates of occurrence.…”

“…If the 2 observers agreed on burrow mound presence or absence, we scored the cell according to their assessment. A. E. Semerdjian, W. T. Bean, or a third observer (A. Rutrough) who had prior experience with this method and the species in the field reviewed observations in which the 2 primary observers disagreed on whether they observed burrow mounds (Rutrough et al 2019). We considered only cells with average confidence scores ≥3 in further analysis.…”

“…For each 1‐km 2 cell, we noted whether at least 2 observers detected giant kangaroo rat burrows, and calculated the mean confidence score from all observers. We used a MaxEnt model designed by Rutrough et al (2019) to quantify habitat suitability throughout our study area. The model used presence points derived from giant kangaroo rat burrow detections in aerial imagery dating before 1960, before the species' range was reduced by agricultural expansion, to model the giant kangaroo rat's fundamental niche (Rutrough et al 2019).…”

“…We used a MaxEnt model designed by Rutrough et al (2019) to quantify habitat suitability throughout our study area. The model used presence points derived from giant kangaroo rat burrow detections in aerial imagery dating before 1960, before the species' range was reduced by agricultural expansion, to model the giant kangaroo rat's fundamental niche (Rutrough et al 2019). Mean annual precipitation, slope, and percent silt were the predictors in their model, which was quite accurate, with a Boyce index of 0.96 calculated from an independent data set.…”

“…Mean annual precipitation, slope, and percent silt were the predictors in their model, which was quite accurate, with a Boyce index of 0.96 calculated from an independent data set. Rutrough et al (2019) projected this model using modern climatic values to estimate current abiotic habitat suitability for giant kangaroo rats. We used estimates from the modern projection to assign binary values of high quality or low quality to each of our 1‐km 2 cells using the maximum sensitivity plus specificity threshold (Bean et al 2012), with values calculated using the PresenceAbsence package in R (Freeman and Moisen 2008, R Core Team 2018).…”