Abstract The American burying beetle (ABB), Nicrophorus americanus (Olivier; Coleoptera: Silphidae), historically occurred in the eastern 35 U.S. States from Canada to Texas and is classified as a habitat generalist. The ABB was listed as a federally endangered species in 1989 with remaining distribution in only six U.S. States. Within these states, populations of ABB are disjunct, occurring in mostly undisturbed habitats associated with multiple soil types and vegetation structure. In Nebraska, the distribution of the ABB has been mapped in two ecoregions, the Sandhills and the Loess Canyons. In this project, we developed and compared a logistic regression model and a random forest model of ABB distribution at its northern and eastern edge in the Northern Plains ecoregions of Nebraska and South Dakota. We used baited pitfall sampling for five trap nights at 482 unique sites to establish presence of ABB at 177 sites. Distribution was not uniform in this ecoregion and the random forest model better predicted occurrence in this area. The results show that the ABB population in the northern plains ecoregion is unique from the previous model of the Nebraska Sandhills despite these ecoregions being adjacent. The model results also reduce requirements to survey and conduct habitat mitigation for ABB in approximately 77,938 hectares of Nebraska and South Dakota that was considered potential habitat while prioritizing areas for conservation. Species distribution models are becoming increasingly important for conservation of rare and endangered species, allowing new populations to be located (e.g., Guisan et al. 2006, Jurzenski et al. 2014), improved conservation area planning, and prediction of potential effects of global climate change (e.g., Carvalho et al. 2011, Riordan and Rundel 2014). Many threatened and endangered species have limited distribution and specific habitat associations; however, some species once had wide distributions that spanned a multitude of habitats (Jones 1963). The use of many habitats may limit the accuracy of predictive models and thus, limit the usefulness of the models for predicting species occurrence and the effects of landscape and climate change on the species. The American burying beetle (ABB), Nicrophorus americanus (Olivier; Coleoptera: Silphidae), is a federally endangered species native to North America (USFWS 2008). The ABB’s range historically extended into 35 U.S. states and three Canadian provinces (Lomolino and Creighton 1996, Bedick et al. 1999). However, the current natural range is limited to areas within six states: Arkansas, Kansas, Nebraska, Oklahoma, Rhode Island, and South Dakota (Godwin and Minich 2005, USFWS 2008), representing a more than 90% reduction from the historic range of the ABB (Lomolino et al. 1995). Within the six states where it still occurs, habitat associations vary and the remaining western populations (South Dakota, Nebraska, Kansas, Oklahoma, and Arkansas) are disjunct both regionally and within the states (USFWS 2016, Leasure and Hoback 2017). The ABB is characterized as a habitat generalist, and despite more than 25 yr of research since its listing, no critical habitat has been designated because of the variability or contradiction found among variables that are strongly linked with ABB occurrence. Several models of predicted occurrence have been developed for Oklahoma and for two Nebraska populations that occur separately in the Loess Canyons and Sandhills (Crawford and Hoagland 2010, McPherron et al. 2012, Jurzenski et al. 2014). Both the Loess Canyons and Sandhills models included validation and produced an area under the curve (AUC) statistic of 0.765 and 0.82, respectively (McPherron et al. 2012, Jurzenski et al. 2014), suggesting high correlation of the models’ ability to predict occurrence of ABB despite differences in predictive variables. Additional modeling of the distribution of western ABB populations by Leasure and Hoback (2017) confirmed differences between habitat associations of ABB in the northern and southern range. However, there are limited data from the most northern areas in Nebraska and southern South Dakota. Although previous work in South Dakota produced both a distribution and population estimate of ABB (Backlund and Marrone 1997, Backlund et al. 2008) this research used variable trap spacing from 0.2 to 3.22 km, a small trap size, rotted beef kidney bait, and longer survey length than standard protocols (USFWS 2014). In this study, ABB in northern Nebraska and southern South Dakota was sampled and positive and negative trap locations were analyzed based on environmental characteristics that were measured within an 800-m radius around each site. We used GIS to map ecologically-relevant characteristics of climate, soils, land cover, and human impacts on the environment. Because our study area is at the extreme northwestern corner of the ABB range, we hypothesized that decreasing annual precipitation would limit the western edge of the distribution and colder winter temperatures would limit the northern distribution. The predicted geographic distribution of the ABB and its correlations with environmental covariates were compared using two modeling approaches, a machine learning algorithm (random forest) and a generalized linear model (logistic regression). Validation data were collected both within the model’s range and outside of the range to test scalability of the model. Methods Field Methods Presence or absence of the ABB was determined at 456 sites in northern Nebraska and southern South Dakota (Fig. 1) from 2005 to 2015 (mostly post-2008) using federally approved bucket-style baited pitfall traps (Bedick et al. 2004). Sites used to generate the models were selected using stratified random sampling methods. Ecoregions in the study area included the Northwestern Great Plains, the Northwestern Glaciated Plains, and a small fragment of Nebraska Sand Hills (USEPA 2006). Sites were spaced at least 1,600 m apart to maintain independence of samples, based on the assumption that the effective sample radius of baited pitfall traps is about 800 m (Leasure and Hoback 2017). There is some empirical support for this effective trap radius (Creighton and Schnell 1998, Leasure et al. 2012, Butler et al. 2013), but the trap radius is likely to be influenced by environmental conditions including wind and precipitation. Five nights of trapping were conducted at each site to minimize the chance of false negatives. Previous studies have estimated the probability of detecting an ABB population with baited pitfall traps to be about 50% for a night of trapping and between 85.7 (±5.3% SE) to 93.7 (±5.1% SE) after five consecutive trap nights (Leasure et al. 2012, Butler et al. 2013). This would result in a false negative rate of about 3–10% across five nights of trapping, and we considered this error rate satisfactory for the purposes of this study (Butler et al. 2013, USFWS 2014). Fig. 1. View largeDownload slide Study area and recovery data for Nicrophorus americanus in Nebraska and South Dakota (2005–2015). Fig. 1. View largeDownload slide Study area and recovery data for Nicrophorus americanus in Nebraska and South Dakota (2005–2015). Trap locations were selected by identifying areas in the Northwestern Great Plains and Glaciated Plains ecoregions that lacked presence/absence sample data during the last 10 yr, were accessible by public roads, and were not within 1,600 m of previously sampled areas. Surveys were conducted using federally compliant 18.9 liters in-ground bucket pitfall traps (USFWS 2014). These traps were dug into the ground with approximately 3 cm of the bucket lip above ground to prevent the entrance of water during rain events. Soil was packed against the outside of the bucket lip to create a ramp to ease the entrance of beetles into the trap. Approximately 8 cm of moistened soil was added to the bottom of the bucket to reduce competition among individuals, and protect against ABBs overheating or desiccating. Soil moisture was checked daily and water was added if needed. Each trap was baited with an extra-large previously frozen laboratory rat carcass (RodentPro.com) which had been rotted in a dark colored 18.9-liter bucket in the sun for 2–4 d, depending on temperature. Traps were covered using two, 5 × 5 cm sticks cut into 45 cm lengths and a piece of plywood measuring 45 × 45 cm. The sticks were placed on the lip of the bucket in parallel to allow beetles space to enter the trap, and the plywood was then placed on top of the sticks. A large piece of sod was then placed on top of the plywood to prevent removal by scavengers or wind. Upon capture of an ABB, the individual was aged, sexed, and its pronotum width measured (USFWS 2014). Because open bait allowed direct contact between the ripened rat carcass and the captured beetle, additional handling and feeding time was not required. Beetles were released approximately 100 m away from traps where they were caught to reduce the likelihood of artificially high recapture rates. At the release site, an artificial burrow was created using a stick, and individual beetles were oriented into the hole into which they readily crawled. A small amount of loose soil or vegetation was then placed over the opening. A single capture of ABB resulted in a positive result for the trap site, while no ABB over five trap nights resulted in a negative result. Environmental Covariates Based on results from previous studies (Leasure and Hoback 2017), we identified 16 environmental covariates to assess as predictors of ABB occurrence including metrics of climate, soil texture, human impacts, and land cover (Table 1). For comparability, an effort was made to use similar predictors to those used in a previous study in the Nebraska Sandhills (Jurzenski et al. 2014). A combination of automated GIS scripts and manual GIS processing was used to delineate an 800-m sample area around each trap location and to summarize the underlying GIS layers representing our covariates within the circular sample areas surrounding each trap location (Python 2012, ESRI 2013). This process was repeated for a grid of points spaced 500 m apart to collect covariate data throughout our study area necessary for mapping the expected distribution of the ABB. Table 1. Predictive variables used in the logistic regression (GLM) and random forest (RF) models to create distribution maps for American burying beetle in northeastern Nebraska and South Dakota GLM RF Covariate Description Citation • Precip Average annual precipitation 1950–2000 Hijmans et al. 2005 • • twinter Avg. min. winter temperature 1950–2000 Hijmans et al. 2005 • • tsummer Avg. summer temperature 1950–2000 Hijmans et al. 2005 • • sand % sand in top soil horizon USDA 2006 • silt % silt in top soil horizon USDA 2006 • clay % clay in top soil horizon USDA 2006 • • road Road density in 2011 (km/ km2) USDC 2011 • hwy Highway density in 2011 (km/ km2) USDC 2011 • • develop % coverage of developed areas Homer et al. 2015 • crop % coverage of crops Homer et al. 2015 • hay % coverage of hayfields Homer et al. 2015 • • water % coverage of open water Homer et al. 2015 • • grass % coverage of grasslands Homer et al. 2015 • • wetgrass % coverage of wet prairies Homer et al. 2015 • • wetland % coverage of wetlands Homer et al. 2015 • • forest % coverage of forests Homer et al. 2015 GLM RF Covariate Description Citation • Precip Average annual precipitation 1950–2000 Hijmans et al. 2005 • • twinter Avg. min. winter temperature 1950–2000 Hijmans et al. 2005 • • tsummer Avg. summer temperature 1950–2000 Hijmans et al. 2005 • • sand % sand in top soil horizon USDA 2006 • silt % silt in top soil horizon USDA 2006 • clay % clay in top soil horizon USDA 2006 • • road Road density in 2011 (km/ km2) USDC 2011 • hwy Highway density in 2011 (km/ km2) USDC 2011 • • develop % coverage of developed areas Homer et al. 2015 • crop % coverage of crops Homer et al. 2015 • hay % coverage of hayfields Homer et al. 2015 • • water % coverage of open water Homer et al. 2015 • • grass % coverage of grasslands Homer et al. 2015 • • wetgrass % coverage of wet prairies Homer et al. 2015 • • wetland % coverage of wetlands Homer et al. 2015 • • forest % coverage of forests Homer et al. 2015 View Large Three climate metrics were selected as environmental covariates: annual precipitation, average minimum winter temperature, and average summer temperature (Table 1). We hypothesized that annual precipitation was related to overall ecosystem productivity and to desiccation risk. Burying beetles are susceptible to desiccation in dry environments leading to increased risk of mortality (Bedick et al. 2006). We also hypothesized that average minimum winter temperature was related to overwintering survival (Schnell et al. 2008), and that average summer temperature influenced habitat suitability as related to temperature-dependent flight activity (Merrick and Smith 2004) during summer months when beetles actively search for carcasses to use for reproduction. Three soil texture covariates were selected: percent sand, silt and clay in the topsoil horizons (O and A) (Table 1). Soil texture has been identified as an important habitat characteristic for determining suitability of areas for ABB to construct underground brood chambers (Scott 1998). In Nebraska, ABB occurrences appear to be related to the presence of sandy loam soils likely because these soils allow rapid burial of carcasses and formation of stable brood chambers, and retain soil moisture (Lomolino, et al. 1995, Scott 1998, Jurzenski et al. 2014). We selected five metrics of human influence: road density, highway density, coverage of developed areas, coverage of crops, and coverage of hayfields (Table 1). These metrics could all have indirect effects on habitat suitability because of general habitat degradation and fragmentation that could affect availability of carcasses suitable for reproduction across the landscape (USFWS 1991, McPherron et al. 2012, Jurzenski et al. 2014). In addition to these indirect effects, intensive row crop agriculture, hayfields, and developed areas could have direct negative effects on the ABB from soil disturbance and pesticide applications. Five land cover metrics were selected: coverage of water, grasslands, wet prairies, wetlands, and forests (Table 1). In dry environments, availability of open water could potentially benefit burying beetles at risk of desiccation, but in general we would expect open water to be negatively related to burying beetle abundance due to decreased availability of terrestrial habitats and potential limits to dispersal. Previous studies have indicated that ABBs were associated with grasslands and wet prairies in Nebraska (Kozol et al. 1988, Bedick et al. 1999, Jurzenski et al. 2014). Analysis Initially, we examined the response of ABB based on presence or absence to the 16 GIS-based environmental covariates within the 800-m buffer surrounding each trap site. We collected data from a total of 456 sites. Covariates were centered and scaled prior to analysis, except percentages that were remained unscaled. We compared results from two modeling approaches, a generalized linear model and a random forest model (Breiman 2001). The generalized linear model (logistic regression) was implemented using the R statistical programming language (R Core Team 2014). To avoid collinearity among predictors in the model, environmental covariates were screened based on Spearman correlation coefficients greater than 0.6. Nine of the 16 environmental covariates were selected for logistic regression (Table 2). Predictors were centered and scaled prior to analysis. Regression coefficients and P-values were used to infer the strength and direction of correlations among ABB occurrence and environmental covariates in the model. The influence of each observation on model parameters (leverage) was assessed graphically using the ‘glm diag plots’ and ‘influence measures’ functions from the R package boot (Canty and Ripley 2014). We focused on the Cook’s D and hat statistics. High leverage observations were removed to avoid a small number of points having a large pull on the predictions produced. Table 2. Regression coefficients for each predictor in the logistic regression model of American burying beetle occurrence in northeastern Nebraska and South Dakota Coefficient Estimate SE P (Intercept) 0.812 0.132 <0.001 tsummer 0.084 0.151 0.579 twinter 1.25 0.146 <0.001 sand 0.07 0.138 0.61 develop −0.041 0.145 0.779 road −0.169 0.151 0.265 grass 0.201 0.15 0.179 wetgrass 0.378 0.146 0.01 water −0.605 0.293 0.039 forest −0.75 0.216 <0.001 Coefficient Estimate SE P (Intercept) 0.812 0.132 <0.001 tsummer 0.084 0.151 0.579 twinter 1.25 0.146 <0.001 sand 0.07 0.138 0.61 develop −0.041 0.145 0.779 road −0.169 0.151 0.265 grass 0.201 0.15 0.179 wetgrass 0.378 0.146 0.01 water −0.605 0.293 0.039 forest −0.75 0.216 <0.001 View Large The random forest model was implemented using the R package ‘randomForest’ (Liaw and Weiner 2002). Random forest is a machine learning algorithm that produces ensemble predictions from a large number of classification trees trained on bootstrap samples of the data. We used 10,000 trees in our model. To build each tree, the algorithm first selected four predictors at random and searched for a threshold that could be applied to one of them to best separate our samples into ABB presence versus absence sites. The best predictor and threshold were retained as the first node in the tree, and the algorithm moved to the next node in each branch (i.e., above and below the selected threshold) to randomly select a new set of four predictors to assess. We included all 16 predictors in the random forest model because the algorithm handles correlations among predictors. Model fit was assessed based on the ‘out-of-bag’ classification error rate. Out-of-bag error rates are produced by the random forest algorithm by making predictions at each site using only those trees in the model that did not include that site in their training data (see Breiman 2001 for more). These error rates are considered conservative estimates that reflect expected error when extrapolating the model to new sites within the study area. The importance of predictors was assessed based on decreases in the Gini index (a measure of homogeneity in predicted presence and absence bins) resulting from random permutations of each predictor (Breiman 2001, Liaw and Wiener 2002). To compare fit between the random forest model and the logistic regression, the AUC statistic was calculated using the R package ROCR (Sing et al. 2005). AUC is a measure of how well predicted probabilities of the ABB occurrence fit our presence-absence observations. It is a threshold-independent fit statistic, meaning that we do not arbitrarily select a threshold for inferring presence or absence based on predicted probabilities of occurrence. We used AUC to compare model fit between our logistic regression and random forest models. AUC greater than 0.8 is considered a good fit (Franklin 2010). We identified a threshold for each model to convert probabilities of occurrence into binary presence-absence predictions that balanced false positive and false negative rates using R package ROCR (Sing et al. 2005). To validate the model and to test the models’ transferability to new regions (both GLM and random forest), we assessed predictive performance at 151 new sites in 2016. We converted predicted probabilities of ABB occurrence to discrete presence-absence predictions using the thresholds identified above that balanced rates of false negative and false positives. Predictor variables for validation sites were collected using the same GIS process described above. Results Of the 456 trap sites, 177 sites were positive, with a total of 1,201 ABB captured (Fig. 1). Both models had relatively good fits to the data within our study area. The logistic regression model had an AUC of 0.83 and a nonsignificant chi-square deviance test (P = 0.529). Five false negative trap sites were removed due to high leverage based on Cook’s D and hat statistics. This is not surprising based on previous studies that have shown five percent or more beetles present in an area cannot be caught even under ideal conditions after five trap nights (Butler et al. 2013). The random forest model had an AUC of 0.82 and an out-of-bag classification error rate of 25.6%. A threshold of 0.4 to convert probabilities of occurrence to discrete presence-absence predictions balanced the false positive and false negative rates for both models. The predicted distributions of ABB from the two models also agreed closely (Fig. 2). Fig. 2. View largeDownload slide Model predictions throughout study area. Color schemes use a probability of occurrence of 0.4 as the threshold to distinguish a presence versus an absence site because this balances the rates of false positives and false negative in both models of ABB probability of occurrence. Fig. 2. View largeDownload slide Model predictions throughout study area. Color schemes use a probability of occurrence of 0.4 as the threshold to distinguish a presence versus an absence site because this balances the rates of false positives and false negative in both models of ABB probability of occurrence. Although both models fit relatively well within our study area, neither model performed well outside of our study area (47% error rate; Fig. 3). The random forest model had far better predictive accuracy than logistic regression within the study area at sites with original training data (Fig. 3). It should be noted that predictions in Fig. 3 are from the full random forest model, whereas AUC for random forest (above) were more conservative assessments based on ‘out-of-bag’ model predictions. Fig. 3. View largeDownload slide Assessment of model prediction accuracy inside and outside the original study area. Predictions inside the study area are for the sites used to fit the model, while predictions outside the study area are new sites. The random forest model had better predictive accuracy within the study area. Both models performed poorly outside the original study area when predicting probability of occurrence of ABB. Fig. 3. View largeDownload slide Assessment of model prediction accuracy inside and outside the original study area. Predictions inside the study area are for the sites used to fit the model, while predictions outside the study area are new sites. The random forest model had better predictive accuracy within the study area. Both models performed poorly outside the original study area when predicting probability of occurrence of ABB. The importance of predictors in each model were noticeably different, except that minimum average winter temperature (twinter) was always a strong predictor. For the random forest model the most important predictors (Fig. 4) were minimum average winter temperature, average precipitation (which was negatively correlated with minimum average winter temperature), clay, which was correlated with sand and silt, grasslands which were negatively correlated with crops, and roads. In the logistic regression, minimum average winter temperature and percent coverage of wet-grasslands had significant positive relationships with the presence of ABB, while the presence of water and forest had significant negative relationships (Table 2, Fig. 5). Fig. 4. View largeDownload slide Importance of predictor variables in the random forest model of American burying beetle occurrence in northeastern Nebraska and South Dakota. Fig. 4. View largeDownload slide Importance of predictor variables in the random forest model of American burying beetle occurrence in northeastern Nebraska and South Dakota. Fig. 5. View largeDownload slide Partial regression plots of significant predictors in the logistic regression model for American burying beetle occurrence in northeastern Nebraska and South Dakota. Fig. 5. View largeDownload slide Partial regression plots of significant predictors in the logistic regression model for American burying beetle occurrence in northeastern Nebraska and South Dakota. Discussion This study represents the first model created specifically for predicting occurrence of ABB in the northwestern limit of its known current range. Our results showed minimum average winter temperature, which was not included in either the Loess Canyons or Sandhills models, as the strongest single predictive factor in both the random forest and linear regression models (McPherron et al. 2012, Jurzenski et al. 2014). The average minimum winter temperature was obtained for the average from 1950 to 2000. Correlation with warmer average winter temperatures may suggest that areas with a lower amount of temperature fluctuation during the winter months increase the likelihood of ABB occurrence. Stable winter temperatures may also allow accumulation of snow in some areas which could act as further insulation for buried beetles. Alternatively, the effects of temperature could be a result of groundwater influence. The Northern Plains lacks large bodies of open water; however, the Ogallala Aquifer is close to the surface (McMahon et al. 2007). The areas with highest likelihood of ABB occurrence in both the random forest and generalized linear models are also close to the surface groundwater from the Ogallala Aquifer (McMahon et al. 2007). This would suggest that minimum average winter temperature may be acting as a surrogate for proximity to subsurface water rather than minimum average winter temperature alone and may relate to avoidance of overwintering desiccation (Bedick et al. 2006). This correlation between ABB occurrence and proximity to subsurface water may also partly explain the differences in habitat association between Nebraska and Oklahoma populations of ABB. Lomolino and Creighton (1996) noted that in its southern range, the ABB appeared to be a forest specialist. The results of this study contradict their finding because ABB had negative correlations with forest and open waters, which were often surrounded by trees in this study region. The correlation may be another product of an association of Nebraska ABBs with the Ogallala Aquifer as a source of soil moisture and temperature stability. ABBs in Oklahoma cannot access aquifer moisture and likely depend on forestation and tree cover to help retain soil moisture (Lomolino and Creighton 1996, Mcmahon et al. 2007, Walker and Hoback 2007). ABBs present in the Loess Canyons were also found to be associated with water features and trees, and in these areas, the Ogallala Aquifer is deep underground and inaccessible to the beetles (Mcmahon et al. 2007, McPherron et al. 2012). Precipitation was a strong predictor of ABB occurrence in the random forest model, which is in agreement with Jurzenski et al. (2014) who found precipitation to be the strongest predictive factor. ABB was negatively associated with clay in the random forest model, which also agrees with the Sandhills model. In the Sandhills, ABBs appear to prefer a more sand dominant soil texture, but will avoid areas without trace amounts of silt and clay as the soil is likely unstable for maintaining a brood chamber (Jurzenski et al. 2014). In contrast, the Loess canyon soil is fine silt and sandy soils are limited in the region (Bedick et al. 1999, McPherron et al. 2012). The Sandhills and Loess canyons differ in suitability for rowcrop agriculture with the Sandhills having unstable or saturated soils that limits crops and the Loess canyons having steep topography. The presence of human development represented by crops and roads were negative predictors in the random forest model, while areas with high percentages of grassland were positive predictors of ABB presence. These findings are consistent with past models and literature that show ABBs to avoid areas of developed land such as agriculture (Sikes and Raithel 2002, McPherron et al. 2012, Jurzenski et al. 2014). As crop values increase, marginal lands are often developed (Lichtenberg 1989) leading to both direct (mortality from pesticides, soil disturbance) and indirect (changes in vertebrate species, habitat fragmentation) impacts on ABB. Future studies should seek to utilize or generate detailed distribution data of not only the ABB but also common carrion sources. Carrion availability is likely the single most important factor in determining presence of ABBs, but due to the lack of fine scale distribution data across a broad range of possible carrion sources, such modeling efforts remain out of reach at the time of this study. While the extinction of the passenger pigeon (Ectopistes migratorius L.) and subsequent loss of suitable carrion source has been widely implicated as a driving force in the endangerment of ABB, the wide-scale suppression and loss of black-tailed prairie dog (Cynomys ludovicianus Ord) towns has not been suggested as a contributing factor to the loss of ABB. Populations of black tailed prairie dogs are currently about 2% of their historical population size (Summers and Linder 1978). Black-tailed prairie dog towns support a rich diversity of vertebrates as potential carrion sources of appropriate size for use by ABBs, but such an association has yet to be properly evaluated (Whicker and Detling 1988, Sikes and Rathel 2002, Lomolino and Smith 2003). Acknowledgments We thank Theresa Andrew, Daniel Snethen, Hurian Gallinari Holzhausen, Márcio Pistore Santos, Alaor Ribeiro da Roca Neto, Gustavo Carvalho Ragazani, and Tiago Corazza da Rosa for assistance in the field and the Oklahoma Agricultural Experiment Station and the Nebraska Department of Roads for funding this project. Dr. Bruce Noden and Scot Stapp provided helpful comments on an earlier version of this manuscript. References Cited Backlund, D. C., and Marrone G. M.. 1997. New records of the Endangered American burying beetle, Nicrophorus americanus Olivier, (Coleoptera: Silphidae) in South Dakota. Coleopt. Bull . 51: 53– 58. Backlund, D. C., Marrone G. M., Williams C. K., and Tilmon K.. 2008. Population estimate of the endangered American burying beetle, Nicrophorus americanus Olivier (Coleoptera: Silphidae) in South Dakota. Coleopt. Bull . 62: 9– 15. Google Scholar CrossRef Search ADS Bedick, J. C., Ratcliffe B. C., Hoback W. W., and Higley L. G.. 1999. Distribution, ecology, and population dynamics of the American burying beetle [Nicrophorus americanus Olivier (Coleoptera: Silphidae)] in south-central Nebraska, USA. J. Insect Conserv . 3: 171– 181. Google Scholar CrossRef Search ADS Bedick, J. C., Ratcliffe B. C., and Higley L. H.. 2004. A New sampling protocol for the endangered American burying beetle, Nicrophorus americanus Olivier (Coleoptera: Silphidae). Coleopt. Bull . 58: 57– 70. Google Scholar CrossRef Search ADS Bedick, J. C., Hoback W. W., and Albrecht M. C.. 2006. High water-loss rates and rapid dehydration in the burying beetle, Nicrophorus marginatus. Physiol. Entomol . 31: 23– 29. Google Scholar CrossRef Search ADS Breiman, L. 2001. Random Forests. Mach Learn 45: 5– 32. Google Scholar CrossRef Search ADS Butler, S. R., Harms R., Farnsworth-Hoback K., Koupal K., Jurzenski J., and Hoback W. W.. 2013. Standardized capture rate of the endangered American burying beetle, Nicrophorus americanus Olivier (Coleoptera: Silphidae) using different trap protocols. J. Insect Conserv . 17: 607– 613. Google Scholar CrossRef Search ADS Canty, A., and Ripley B.. 2014. Boot: Bootstrap R (S-Plus) Functions. R package version 1.3–13 . Carvalho, S. B., Brito J. C., Crespo E. G., Watts M. E., Possingham H. P.. 2011. Conservation planning under climate change: toward accounting for uncertainty in predicted species distributions to increase confidence in conservation investments in space and time. Biol. Conserv . 144: 2020– 2030. Google Scholar CrossRef Search ADS Crawford, P. H. C., and Hoagland B. W.. 2010. Using species distribution models to guide conservation at the state level: the endangered American burying beetle (Nicrophorus americanus) in Oklahoma. J. Insect Conserv . 14: 511– 521. Google Scholar CrossRef Search ADS Creighton, C. J. and Schnell G. D.. 1998. Short-term movement patterns of the endangered American burying beetle Nicrophorus americanus. Biol. Conserv . 86: 281– 287. Google Scholar CrossRef Search ADS ESRI. 2013. ArcGIS 10.2 for Desktop. Version 10.2.0.3348 . Esri Inc., Redlands, CA. Franklin, J. 2010. Mapping species distributions: spatial inference and prediction . Cambridge University Press, Cambridge, UK. Google Scholar CrossRef Search ADS Godwin, W. B., and Minich V.. 2005. Status of the American burying beetles, Nicrophorus americanus Olivier, (Coleoptera: Silphidae) at Camp Maxey, Lamar county, Texas. Interagency Final Report to Texas Army National Guard . Stephen F. Austin State University, Nacogdoches, TX. Guisan, A., Broennimann O., Engler R., Yoccoz N. G., Vust M., Zimmermann N. E.. 2006. Using niche-based models to improve the sampling of rare species. Conserv. Biol . 20: 501– 511 Google Scholar CrossRef Search ADS PubMed Hijmans, R. J., Cameron S. E., Parra J. L., Jones P. G., and Jarvis A.. 2005. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol . 25: 1965– 1978. http://www.worldclim.org/current Google Scholar CrossRef Search ADS Homer, C. G., Dewitz J. A., Yang L., Jin S., Danielson P., Xian G., Coulston J., Herold N. D., Wickham J. D., and Megown K.. 2015. Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information. Photogramm Eng Remote Sensing 81: 345– 354. http://www.mrlc.gov/nlcd2011.php Jones, R. E. 1963. Identification and analysis of lesser and greater prairie chicken habitat. J. Wildl. Manag . 27: 757– 778. Google Scholar CrossRef Search ADS Jurzenski, J. D., Jorgensen C. F., Bishop A., Grosse R., Riens J., and Hoback W. W.. 2014. Identifying priority conservation areas for the American burying beetle, Nicrophorus americanus (Coleoptera: Silphidae), a habitat generalist. Systematics and Biodiversity 12: 149– 162. Google Scholar CrossRef Search ADS Kozol, A. J., Scott M. P., and Traniello J. F. A.. 1988. The American burying beetle, Nicrophorus americanus: studies on the natural history of a declining species. Psyche 95: 167– 176. Google Scholar CrossRef Search ADS Leasure, D. R. and Hoback W. W.. 2017. Distribution and habitat of endangered American burying beetle in northern and southern regions. J. Insect Conserv 21: 75– 86. doi: 10.1007/s10841-017-9955-5 Google Scholar CrossRef Search ADS Leasure, D. R., Rupe D. M., Phillips E. A., Opine D. R., and Huxel G. R.. 2012. Efficient new above-ground bucket traps produce comparable data to that of standard transects for endangered American burying beetles (Silphidae: Nicrophorus americanus Olivier). Coleopt. Bull . 66: 209– 218. Google Scholar CrossRef Search ADS Liaw, A., and Weiner M.. 2002. Classification and Regression by randomForest. R News 2: 18– 22. Lichtenberg, E. 1989. Land quality, irrigation development, and cropping patterns in the northern high plains. Am. J. Agric. Econ . 71: 187– 194. Google Scholar CrossRef Search ADS Lomolino, M. V., and Creighton J. C.. 1996. Habitat selection, breeding success and conservation of the endangered American burying beetle Nicrophorus americanus. Biol. Conserv . 77: 235– 241. Google Scholar CrossRef Search ADS Lomolino, M. V., and Smith G. A.. 2003. Prairie dog towns as islands: applications of island biogeography and landscape ecology for conserving non-volant terrestrial vertebrates. Glob. Ecol. Biogeogr . 12: 275– 286. Google Scholar CrossRef Search ADS Lomolino, M. V., Creighton J. C., Schnell G. D., and Certain D. L.. 1995. Ecology and conservation of the endangered American burying beetle (Nicrophorus americanus). Conserv. Biol . 9: 605– 614. Google Scholar CrossRef Search ADS McMahon, P. B., Dennehy F. K., Bruce B. W., Gurdak J. J., and Qi S. L.. 2007. Water-Quality Assessment of the High Plains Aquifer, 1999–2004. United States Geological Survey: National Water-Quality Assessment Program, professional paper 1749. https://pubs.usgs.gov/pp/1749/ McPherron, M. M., Jurzenski J., Albrecht M. A., Farnsworth-Hoback K. M., Walker T. L., and Hoback W. W.. 2012. Model of habitat suitability for American burying beetles in Nebraska’s Loess Canyons ecosystem. Trends in Entomology 8: 27– 36. Merrick, M. J., and Smith R. J.. 2004. Temperature regulation in burying beetles (Nicrophorus spp.: Coleoptera: Silphidae): effects of body size, morphology, and environmental temperature. J. Exp. Biol . 207: 723– 733. Google Scholar CrossRef Search ADS PubMed Python. 2012. Python v2.7.3. Python Software Foundation. http://www.python.org. R Development Core Team. 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org. Riordan, E. C., and Rundel P. W.. 2014. Land use compounds habitat losses under projected climate change in a threatened California ecosystem. PLoS One 9: e86487. Google Scholar CrossRef Search ADS PubMed Schnell, G. D., Hiott A. E., Creighton J. C., Smyth V. L., and Komendat A.. 2008. Factors affecting overwinter survival of the American burying beetle, Nicrophorus americanus, (Coleoptera: Silphidae). J. Insect Conserv . 12: 483– 492. Google Scholar CrossRef Search ADS Scott, M. P. 1998. The ecology and behavior of burying beetles. Annu. Rev. Entomol . 43: 595– 618. Google Scholar CrossRef Search ADS PubMed Sikes, D. S., and Raithel C. J.. 2002. A review of hypotheses of decline of the endangered American burying beetle (Silphidae: Nicrophorus americanus Olivier). J. Insect Conserv . 6: 103– 113. Google Scholar CrossRef Search ADS Sing, R., Sander O., Beerenwinkel N., and Lengauer T.. 2005. ROCR: Visualizing classifier performance in R. Bioinformatics 21: 7881. Google Scholar CrossRef Search ADS Summers, A., and Linder R. L.. 1978. Food habits of the black-tailed prairie dog in western South Dakota. J. Range Manage . 31: 134– 136. Google Scholar CrossRef Search ADS (USDA) U.S. Department of Agriculture. 2006. Digital General Soil Map of U.S. United States Department of Agriculture, Natural Resources Conservation Service, Fort Worth, TX. http://catalog.data.gov/harvest/object/69d5a4d1-a4ef-4208-bf4b-898c0b67a425/html (USDC) U.S. Department of Commerce. 2011. TIGER/Line Shapefile, 2011, State, Arkansas, Primary and Secondary Roads State-based Shapefile. U.S. Department of Commerce, U.S. Census Bureau, Geography Division. http://www2.census.gov/geo/tiger/TIGER2011/PRISECROADS/tl_2011_05_prisecroads.zip (USEPA) U.S. Environmental Protection Agency. 2006. Ecoregions of North America. Level III Ecoregions. https://www.epa.gov/eco-research/ecoregions-north-america (USFWS) U.S. Fish and Wildlife Service. 1991. American Burying Beetle (Nicrophorus americanus) Recovery Plan . Newton Corner, MA. (USFWS) U.S. Fish and Wildlife Service. 2008. American burying beetle (Nicrophorus americanus) 5 year review: Summary and Evaluation . New England Field Office, Concord, NH. (USFWS) U.S. Fish and Wildlife Service. 2014. American burying beetle Nicrophorus americanus Oklahoma Presence/Absence Live-trapping Survey Guidance . United States Department of Interior, Fish and Wildlife Service, Division of Ecological Services, Tulsa, OK. (USFWS) U.S. Fish and Wildlife Service. 2016. American burying beetle: Additional information. https://www.fws.gov/southwest/es/oklahoma/ABB_Add_Info.htm. Walker, T. L., and Hoback W. W.. 2007. Effects of invasive eastern redcedar on capture rates of Nicrophorus americanus and other Silphidae. Environ. Entomol . 36: 297– 307. Google Scholar CrossRef Search ADS PubMed Whicker, A. D., and Detling J. K.. 1988. Ecological consequences of prairie dog disturbances. Bioscience 38: 778– 785. Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of Entomological Society of America. All rights reserved. For permissions, please e-mail: email@example.com.
Insect Systematics and Diversity – Oxford University Press
Published: Jan 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera