Printer Friendly

Habitat characteristics within a zone of separation between the ranges of two species of pocket gophers.

Pocket gophers (family Geomyidae) are subterranean rodents found throughout North America. Their solitary lifestyles, limited dispersal capabilities, and similar habitat requirements makes them ideal models for investigating patterns of distributions and ecological interactions. Species of pocket gophers generally are not sympatric in their distributions and when their distributions overlap, they tend to separate themselves ecologically. This distributional pattern is commonly referred to as parapatry (Key, 1982; Bull, 1991). Several studies have identified patterns of parapatry among pocket gophers (Miller, 1967; Vaughan, 1967; Thaeler, 1968). Specifically the distributional patterns of the plains pocket gopher (Geomys bursarius) and yellow-faced pocket gopher (Cratogeomys castanops) have received much attention where the geographic ranges of these species overlap primarily in portions of New Mexico, Colorado, and Kansas (Hall, 1981). In these areas of overlap, both species separate themselves by inhabiting areas with different soil types and land cover (Best, 1973; Moulton et al., 1983; Lovell et al., 2004; Hoffman and Choate, 2008).

Although populations of G. bursarius and C. castanops do not co- exist in the same habitats, the spatial characteristics of the zones of contact between the two species appear to be highly variable across the landscape. In some cases the distributions of each species do not contact each other but rather a gap is present between the populations. Hoffman and Choate (2008) noted a gap existed between the distributions of G. bursarius and C. castanops in certain areas of southwestern Kansas. This phenomenon has also been observed among other species of pocket gophers. Kennerly (1959) found a 17.7 km gap between populations of the Texas pocketgopher (Geomyspersonatus) and G. bursarius. Likewise, Best (1973) documented a 29 km gap between populations of Botta's pocket gopher (Thomomys bottae) and C. castanops. Because these distances are beyond the normal limits of pocket dispersal (Vaughan, 1963; Smolen et al., 1980; Daly and Patton, 1986), any contact among these populations seems unlikely. Although these gaps between populations of pocket gophers appear to be a common occurrence, they are not indicative of all pocket gopher distributions. Both Kennedy (1959) and Best (1973) noted several instances where the occurrences of different species of pocket gophers approached within less than 100 m. In southeastern Colorado Moulton et al., (1979) trapped C. castanops and T. bottae from mounds 20 m apart and noted instances where the distributions of each species overlapped by approximately 800 m. This suggests the limiting factors of pocket gopher distributions are variable across the landscape.

There has been little empirical research conducted to explain the presence of gaps between populations of closely related species. In addition to pocket gophers, several other species including woodrats (Bimey, 1973), chickadees (Brewer, 1963), and warblers (Terborgh, 1971) have exhibited similar patterns. Some of the first work on this topic was a theoretical investigation by MacArthur (1972); he described a graphical situation in which the occurrence of two species was dependent on the abundance of different resources. MacArthur (1972) calculated that in an unproductive series of environments, where the abundance of the two resources was minimal, a gap between the species could exist because of the lack of resources needed to support either. Slade and Robertson (1977) built upon this model by predicting the gap between species could be competitively induced if there was a decrease in resources resulting from competition. Using a fitness-model Cody (1974) showed where two species of warblers existed along a resource gradient and when high proportions of their preferred habitat were required, a hiatus of ranges could result in the middle of the gradient.

Historically, quantifying habitat suitability for a species in areas where the species did not exist was difficult. This can be especially problematic for pocket gophers whose main habitat requirements consist of different soil textures. Ecotonal changes in soil types typically are not as apparent as changes in vegetation. However, with the recent advancements in species distribution modeling, the habitat suitability for a species can be predicted over multiple spatial scales. Species distribution modeling examines relationships between species occurrence and the environment (Segurado and Araujo, 2004; Austin, 2007). The resulting maps produced by these models allow investigators to identify areas of varying habitat quality throughout the landscape. They also can determine the importance of environmental variables in predicting the species distribution. These characteristics make species distribution modeling an excellent tool for studying the parapatric boundaries of species.

The goal of this research was to characterize the habitat conditions within a zone of separation between G. bursarius and C. castanops. Presently no research has quantified habitat suitability for pocket gophers along a parapatric boundary. Following the results of MacArthur (1972) and Cody (1974), it is expected habitat conditions would be less suitable within the zone of separation compared to locations where both species occur. Most research on habitat selection for each species suggests soil characteristics and land cover/ land use are the most important variables for determining pocket gopher presence. For instance G. bursarius is mostly commonly found in coarse-grained sandy soils and cultivated or otherwise disturbed habitats, whereas C. castanops is found in undisturbed grasslands and fine-grained soils such as clay and silt (Downhower and Hall 1966; Moulton et al., 1983; Hoffman and Choate, 2008). The specific objectives of this research were to: (1) describe the characteristics of a zone of separation between G. bursarius and C. castanops, and (2) determine if differences in habitat conditions exist within the zone of separation compared to locations where each species of pocket gopher occur.

MATERIALS AND METHODS

This study was conducted in a nine county area (38.793N, -99.518Wby 37.477N,-101.335W) in southwestern Kansas (Fig. 1). The distribution of G. bursarius and C. castanops in the study area was determined through field surveys conducted in 2002. Pocket gophers were collected using Macabee traps set in fresh pocket gopher mounds located visually by driving slowly on all county roads in the study area. To supplement the occurrence localities obtained through field surveys, a list was compiled of all known museum localities in the study area where specimens of the two species had been collected. The specimens are housed in the collections of the Sternberg Museum of Natural History at Fort Hays State University (MHP) and the Kansas University Natural History Museum (KU). All specimens collected during fieldwork were prepared as vouchers and deposited in MHP. A list of specimens examined can be found in Hoffman and Choate (2008).

To characterize the habitat conditions within the study area, digitized maps of soil permeability, porosity, rock fragment volume, texture (% sand, % silt, % clay), and land cover were downloaded into a GIS. All soil data were downloaded from the Soil Information for Environmental Modeling and Ecosystem Management (http://www.soilinfo.psu.edu). The map of land cover was derived from 1992-1993 data, collected from the Advanced Very High Resolution Radiometer (AVHRR) and consists of 12 land cover classifications (Hansen et al., 2000). All environmental data possessed a 1 km spatial resolution. To account for colinearity, a correlation matrix of all continuous variables was calculated for each species. When two variables possessed a correlation value >0.5, only one of those variables was chosen for the analysis of habitat quality. The decision on which correlated variable to include in the model was based upon the literature and knowledge of pocket gopher biology. For instance percent sand was highly correlated with percent silt and several studies have documented the importance of percent sand for influencing pocket gopher distributions over percent silt (Moulton et al., 1983; Lovell et al., 2004; Hoffman and Choate, 2008). Therefore, percent sand was included in the model and percent silt was discarded. Because land cover was a categorical variable, and therefore not correlated with any other variable, it was included in the model for both species.

Habitat suitability within the study area was determined for both species using the predictive habitat model, Maxent (3.3.2). All of the uncorrelated environmental variables were included into a single model to produce maps of habitat suitability. The Maxent algorithm operates on a set of constraints that describe what is known from the sample of the target distribution. Maxent does not require information on where a species does not occur (absence localities). Rather it characterizes the background environment with a sample of background points. Species occurrence at these background points is unknown (Phillips et al, 2006). The program generated 10,000 random locations in the study area to use as background points. Maxent predicts the probability distribution across the study area and employs maximum entropy principles and regularization parameters to prevent over-fitting and is capable of building nonlinear response curves using different feature classes (Phillips et al., 2006; Phillips and Dudik, 2008). A variety of feature classes were used including linear, quadratic, threshold, hinge, and auto feature, in order to maximize model fit. Maxent has become a popular method and performs well when compared to other presence-only and presence-absence models (Elith et al., 2006; Phillips et al., 2006; Hoffman et al., 2010). Habitat suitability maps were produced using the logistic output in Maxent for which the probability of occurrence ranged from 0-1. Finally, species response curves were calculated in the Maxent algorithm that shows how probability of occurrence changes as the environmental variable is varied while keeping all other variables at the average sample value.

The accuracy of the model predictions was determined in multiple ways. The first means of evaluation included calculating area under the receiver operating characteristic curve (AUC) values. The AUC curve is a plot of the sensitivity vs. 1-specificity at all possible threshold probabilities for a positive prediction. The AUC values range, from zero to one, with those close to unity indicating better predictive power. The AUC values are interpreted as the probability that a randomly chosen pair of occupied and unoccupied sites is correctly predicted. The training and testing AUC values was performed by directing Maxent to build the predicted model using 70% of the occurrence locations (training data) and then use the remaining 30% to test the accuracy of the prediction (testing data). Models with AUC values greater than 0.8 are considered good predictors of species distributions (Swets, 1988; Fielding and Bell, 1997). Secondly, the performance of the model prediction was evaluated by assessing the omission rates of the data using two threshold-dependent approaches provided by Maxent. Models that provide a good prediction should have low omission rates on the data. The first threshold-dependent approach used was the minimum training presence (MTP) threshold. This approach sets the threshold at the lowest value of prediction for any of the locations in the dataset. Therefore, the threshold value used will vary by model. With this approach a binary prediction is produced including all pixels where the suitability is at least equal to the locations where the species is known to be present. The second threshold-dependent approach used was a fixed 10% cumulative probability threshold to find the extrinsic omission rate. In this approach the omission rate for presence localities is expected to equal the threshold employed. These two threshold-dependent approaches have been commonly used to evaluate the performance of Maxent models (Riordan and Rundel, 2009; Anderson and Gonzalez, 2011). A one-tailed binomial probability was used to determine whether the models had omission rates better than a random prediction (Phillips et al, 2006).

A goal of this study was to determine the importance of each environmental variable used to predict the distribution of G. bursarius and C. castanops. In the Maxent algorithm, one measure of variable importance is determined through a jackknife analysis (Phillips et al, 2006). Jackknife analysis of variable importance is determined by calculating the gain of a model built using each variable individually. For each iteration of the Maxent algorithm, the increase or decrease of the regularized gain is either added to or subtracted from the overall contribution of the variable. The jackknife analysis is performed using training and testing gain as well as AUC values on the test data. Training gain was calculated using the 70% training data while testing gain and AUC were calculated using the 30% testing data. For each variable the training gain, the testing gain, and AUC values were reported for a model built with only that variable. Variables with higher values of gain and AUC were considered most indicative of species occurrence.

Habitat conditions and suitability at the presence localities of each species were compared to the conditions within the zone of separation. The zone of separation was determined by drawing a shortest distance line between adjacent presence localities in an area where the distributions of each species approached each other in Finney, Lane, and Ness counties. The predominant vegetation types in this area were characteristic of the adjacent areas and included a mix of native short-grass prairie, cropland, and open shrubland. Presence localities were overlaid onto the digitized maps of environmental variables, and their values were extracted. Only values determined to be most important by the Maxent algorithm in predicting each species occurrence were extracted. To sample the area inside the zone of separation, 100 points within the zone were randomly generated, and the data from the important environmental variable (s) were extracted to those points. In addition to the environmental data, the respective species localities and random points generated within the zone of separation were overlaid onto the habitat suitability maps generated by Maxent and suitability values to those points were extracted. Student 1-tests were used to determine if habitat suitability and habitat conditions were different within the zone of separation compared to presence localities. All statistical analyses were performed using R (2.9.2) statistical software (R Development Core Team, 2008).

RESULTS

The distribution of each species and location of the zone of separation is shown in Figure 1. The narrowest point separating G. bursarius and C. castanops was approximately 9 km and the widest point measured approximately 24 km. Repeated surveys of this area yielded no evidence of pocket gopher activity. Maxent produced good predictions of habitat suitability with both species where both models had training and testing AUC values greater than 0.8 (Table 1). Additional support for the model fit was indicated by the significant results of the omission rates for the two threshold values (Table 1). The fact both omission rates were significant indicates the models prediction was better than a random prediction. The most suitable habitat for G. bursarius was in the southern portion of the study area along the Arkansas River (Fig. 2). Other areas of high suitability for G. bursarius exist north of the river but are smaller and more fragmented. Suitable habitat for C. castanops was found in upland areas north of the river (Fig. 2). This region consists mostly of native uncultivated grasslands primarily used for livestock grazing.

Habitat suitability was lower in the zone of separation for both species. The average habitat suitability at presence locations of G. bursairus (0.54, SD = 0.2) was significantly higher ([t.sub.61] = 9.08, P < 0.0001) than within the zone of separation (0.28, SD = 0.08). Similar results were seen with C. castanops were habitat suitability at collection localities (0.60, SD = 0.31) was significantly higher ([t.sub.70] = 4.25, P < 0.0001) than within the zone of separation (0.34, SD = 0.33).

The jackknife analysis of variable importance showed differences between the two species (Fig. 3). For G. bursarius percent sand showed the most importance in predicting habitat suitability compared to all the other uncorrelated variables. Both training and testing gain along with AUC values were higher than for all the other variables. The average percent sand at the presence localities of G. bursarius (43.4, SD = 27.1) was significandy higher ([t.sub.57] = 6.72, P < 0.0001) than percent sand within the zone of separation (18.8, SD = 4.3). For C. castanops several variables were shown to be important for predicting habitat suitability. All variables possessed relatively high AUC values when used in isolation to predict habitat suitability. Training gain was comparable across all variables except for porosity. Testing gain, however, was noticeably higher for percent clay and sand. All combined, percent clay and sand are the most important variables for predicting the habitat suitability of C. castanops compared to all other uncorrelated variables. At the presence localities of C. castanops, the average percent clay (20.2, SD = 3.03) was significantly higher ([t.sub.104] = 2.99, P = 0.002) than within the zone of separation (18.4, SD = 3.91). Similarly average percent sand (21.5, SD = 4.26) was significantly higher ([t.sub.85] = 3.46, P < 0.001) at presence localities than within the zone of separation (18.8, SD = 4.3).

The species response curves show how predicted probability of occurrences varies with certain soil characteristics. For G. bursarius, percent sand was determined to be the most important variable for predicting occurrence. The response curve for G. bursarius shows probability of occurrence peaks when soils have approximately 43% sand content and remained relatively constant for higher sand percentages (Fig. 4). The average percent sand content of the soil where G. bursarius occurred was 43.4%, which corresponds to a 68% probability of occurrence. Random locations within the zone of separation possessed a much lower percent sand content (18.8%) and probability of occurrence (22%). The response curves for C. castanops shows a noticeable difference from G. bursarius (Fig. 5). The two soil variables most important for predicting the occurrence of C. castanops were percent sand and clay. Probability of occurrence rose sharply at lower sand contents and peaked at approximately 24% sand. Beyond that value the probability of occurrence declined gradually until reaching zero at 70%. Average percent sand at C. castanops locations was 21.5% which corresponds to a 61% probability of occurrence while percent sand at locations in the zone of separation had an average 18.8% and a 45% probability of occurrence. The response curve for percent clay showed a peak probability of occurrence at soils with 24% clay content. The average percent clay at presence locations of C. castanops was 20.2% and 18.4% at locations within the zone of separation. The predicted probabilities for these sites were 57% and 44%, respectively.

DISCUSSION

There have been multiple studies on the biogeography of pocket gophers. Most note that when the ranges of two species overlap segregation often occurs between the populations (Vaughan, 1967; Thaeler, 1968; Lovell et al., 2004). The results of this study support the hypothesis by MacArthur (1972) that a decreasing resource gradient for two species can result in an area of low habitat suitability between the populations causing a gap to exist. The habitat suitability maps produced by Maxent showed habitat conditions were less suitable in the zone of separation compared to presence locations for both G. bursarius and C. castanops. This was also the case for the soil variables determined to be most important for predicting habitat suitability. Percent sand was the most important variable to G. bursarius. The zone of separation was characterized by significantly less sandy soils than in locations where G. bursarius was captured. For C. castanops percent sand and clay were most important for predicting habitat suitability. Soils found in the zone of separation consisted of significantly less percent sand and clay compared to presence locations of C. castanops. It is not surprising the importance of percent sand and clay differed between G. bursarius and C. castanops. For instance G. bursarius consistently has been shown to prefer sandy soils while C. castanops tends to inhabit a wider range of soil textures (Hoffman and Choate, 2008; Moulton et al, 1983; Lovell et al., 2004). In a similar study, which included some of the same populations of pocket gophers, Hoffman et al. (2007) identified large contiguous tracts of land where no pocket gophers occurred. These areas were much larger than the zone of separation described in this study, often times spanning several hundred kilometers. They noted soil conditions were also different between presence locations and areas where no pocket gophers occurred but also that land use was significantly different. In these large unoccupied areas, cultivation was prominent and a primary reason for the exclusion of pocket gophers (Hoffman el al., 2007). Conversely, for the zone of separation analyzed in this study, land use was not important for maintaining separation between pocket gophers. This suggests the removal of suitable vegetation through agricultural practices can cause greater separation between pocket gopher populations than soil conditions considered alone.

The manner in which G. bursarius and C. castanops respond to variation in soil texture provides insight into how the zone of separation is maintained. For G. busarius the species response curve showed when percent sand fell below 38% the probability of occurrence began decreasing (Fig. 4). Soils in the zone of separation possessed an average of 18.8% sand which corresponds to only a 21% probability of occurrence. The response of C. castanops to percent sand and clay was noticeably different compared to G. bursairus. For percent sand the response curve is right skewed with a mode at approximately 20% (Fig. 5). The average percent sand found in soils where C. castanops was captured was approximately 21%. As percent sand decreased the probability of occurrence decreased at a faster rate compared to increasing percentages. Soils in the zone of separation possessed an average of 18.8% which corresponded to a 52% of occurrence for C. castanops. The response curve for percent clay assumed a more normal bell shaped curve again with the mode at approximately 20% (Fig. 5). The zone of separation contained soils with an average of 18.5% clay. Soils with this amount of clay predicted a 55% probability of occurrence for C. castanops. The fact there was only a minimal decrease in probability of occurrence for C. castanops when comparing soil textures found at presence locations and the zone of separation suggests that while slight variation in one of the soils variables may not necessarily exclude C. castanops from an area, slight variations in both soil textures is sufficient to make an area unsuitable.

Pocket gopher preference towards certain soil textures is best illustrated in both anatomical and behavioral specializations (Lessa and Thaeler, 1989). Pocket gophers in the genus Geomys possess larger forelimb muscles and front claws compared to Cratogeomys spp. making Geomys more adept for digging primarily with its claws. In contrast species of Cratogeomys have more procumbent incisors and larger jaw muscles which allow for greater utilization of its teeth while digging. When presented with compact soils Cratogeomys consistently will use its teeth to loosen soil while Geomys will only use its claws (Lessa, 1987). The inability to use its teeth to loosen soils could prevent G. bursairus from occupying the more compact soils found in the zone of separation. Explaining the absence of C. castanops in the zone of separation is less apparent because this species seems to have the anatomical specializations for inhabiting the soils found in this region. It is possible C. castanops is more sensitive to changes in multiple soil textures and, therefore, it can only occupy certain soil types. This point is illustrated by the bell-shaped response curves for percent sand and clay where the highest probability of occurrence is approximately 24-25% for each variable (Fig. 5). As the percentages for each of these soil textures increases or decreases, the probability of occurrence decreases for C. castanops. Therefore, even small variations in percent sand and clay can result in an additive decrease in probability of occurrence. This is in contrast to G. bursarius where only one variable (percent sand) was considered important and the response curve assumed a sigmoidal shape (Fig. 4). Probability of occurrence continually increased with increasing amounts of percent sand (albeit at different rates) until it reached approximately 38% sand. At this point the probability of occurrence remains relatively constant for higher percentages of sand.

Parapatry can be separated into hybridization parapatry, where the contacting taxa form a narrow hybrid zone, and ecological parapatry, where no hybridization occurs (Key, 1982). There is no evidence of G. bursarius and C. castanops hybridizing; therefore, this situation would represent a form of ecological parapatry. Bull (1991) summarized the mechanisms most commonly responsible for maintaining parapatric boundaries of nonhybridizing species which included ecotonal changes, interspecific competition, predation, parasites and disease, and reproductive interference. This study specifically addresses the change in soil and landcover ecotones between the two populations with the results suggesting soil conditions are significantly different in the zone of separation for both species. This situation closely resembles a descriptive model proposed by Bull (1991) where populations exist in ridges of favorable habitat, and as they approach each other a trough of lesser quality habitat maintains the parapatric boundary. Trough populations are only maintained if dispersal from ridge populations exceeds mortality. However if the trough is small and/or gradual, so much so that a species can easily disperse through it, then some form of interspecific interaction likely is needed to maintain the parapartric boundary (States, 1976; Bull and Possingham, 1995). Competition between species of pocket gophers has been suggested as a factor explaining their distributions (Miller, 1964; Vaughan, 1967) and interspecific competition can be responsible for maintaining parapatric boundaries (Neet and Hausser, 1990; Heller, 1971; Sheppard, 1971), although there is some disagreement on how important competition is to maintaining parapatric boundaries (Garcia-Ramos et al., 2000; Key, 1982). While it is difficult to confirm, competitive interactions could be present in some of the populations of gophers in this study. For instance the predictive model shows suitable habitat for C. castanops in the middle of the zone of separation and extending into areas occupied by G. bursarius (Fig. 2). It is possible G. bursarius could be competitively excluding C. castanops from these areas but such interactions may not be indicative for most populations given the distance between them. The width of the zone of separation between C. castanops and G. bursarius ranges from 9 km to 24 km. Information on pocket gopher dispersal distances is sparse but suggests most above ground dispersal is by juveniles or subadults and ranges from 40-700 m (Daly and Patton, 1986; Vaughan, 1963; Smolen et al., 1980; Williams and Cameron, 1984). Given their limited dispersal capabilities and the width of unfavorable habitat separating the populations, it is unlikely interspecific interactions are commonplace. However, pocket gophers generally exhibit small-scale aggregations in their distributions. These small populations are unstable, especially in marginal habitat, making the true history of contact difficult to determine. This study is unique compared to previous investigations because habitat suitability was quantified over the entire study area and not just at sites where pocket gophers were captured. Being able to visualize variation in habitat suitability can provide insight into how different species are spatially arranged across the landscape.

Acknowledgments.--I thank J. R. Choate (deceased), R. B. Channell, W. J. Stark, E. J. Finck, and R. J. Zakrewski for their comments, suggestions, and help in designing this project. This work was funded in part by the Department of Biological Sciences, Fort Hays State University. Voucher specimen locations were provided by the Sternberg Museum of Natural History (MHP) and Kansas University Natural History Museum (KU).

SUBMITTED 15 JANUARY 2014

ACCEPTED 26 AUGUST 2014

LITERATURE CITED

Anderson, R. P. and I. Gonzalez, Jr. 2011. Species-specific tuning increases robustness to sampling bias in models of species distributions: an implementation with Maxent. Ecol. Model., 222:2796-2811.

Austin, M. P. 2007. Species distribution models and ecological theory: a critical assessment and some possible new approaches. Ecol. Model., 200:1-19.

Best, T. L. 1973. Ecological separation of three genera of pocket gophers (Geomyidae). Ecology, 54:1311-1319.

Birney, E. C. 1973. Systematics of three species of woodrats (Genus Neotoma) in central North America. Misc. Publ. Univ. Kansas Mus. Nat. Hist., 58:1-173.

Brewer, R. 1963. Ecological and reproductive relationships of black-capped and Carolina chickadees. Auk, 80:9-47.

Bull, C. M. 1991. Ecology of parapatric distributions. Annu. Rev. Ecol. Syst., 22:19-36.

-- and H. Possingham. 1995. A model to explain ecological parapatry. Am. Nat., 145:935-947.

Cody, M. L. 1974. Opdmization in ecology. Science, 183:1156-1164.

Daly, J. C. and J. L. Patton. 1986. Growth, reproduction, and sexual dimorphism in Thomomys bottae pocket gophers. J. Mammal., 67:256-265.

Downhower, J. F. and E. R. Hall. 1966. The pocket gopher in Kansas. Misc. Publ. Univ. Kansas Mus. Nat. Hist., 44:1-32.

Elith, J., C. H. Graham, R. P. Anderson, M. Dudik, S. Ferrier, A. Guisan, R. J. Hijmans, F. Huettmann, J. R. Leathwick, A. Lehmann, J. Li, L. G. Lohmann, B. A. Loiselle, G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, J. McC. Overton, A. T. Peterson, S. J. Phillips, K. Richardson, R. Scachetti-Pereira, R. E. Schapire, J. Soberon, S. Williams, M. S. Wisz, and N. E. Zimmermann. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography, 29:129-151.

Fielding, A. H. and J. F. Bell. 1997. A review of methods for the assessment of prediction errors in conservation presence-absence models. Environ. Conseru., 24:38-49.

Garcia-Ramos, G., F. Sanchez-Garduno, and P. K Maine. 2000. Dispersal can sharpen parapatric boundaries on a spatially varying environment. Ecology, 81:749-760.

Kennerly, T. E. 1959. Contact between the ranges of two allopatric species of pocket gophers. Evolution, 13:247-263.

Key, K H. L. 1982. Species, parapatry, and the morabine grasshoppers. Syst. Zool., 30:425-458.

Hall, E. R. 1981. Mammals of North America, 2nd ed. John Wiley and Sons, New York, NY.

Hansen, M. C., R. S. Defries, J. R. G. Townshend, and R. Sohlberg. 2000. Global land cover classification at 1 km spatial resolution using a classification tree approach. Int. J. Remote. Sens., 21:1331-1364.

Heller, H. C. 1971. Altitudinal zonation of chipmunks (Eutamias): interspecific aggression. Ecology, 52:312-319.

Hoffman, J. D., N. N. Aguilar-Amuchastegui, and A. J. Tyre. 2010. Use of simulated data from a process-based habitat model to evaluate methods for predicting species occurrence. Ecography, 33:656-666.

-- and J. R. Choate. 2008. Distribution and status of the yellow-faced pocket gopher in Kansas. West. N. Am. Nat., 67:483-492.

--, --, and R. Channell. 2007. Effects of land use and soil texture on distributions of pocket gophers in Kansas. Southwest. Nat., 52:296-301.

Lovell, D. C., W. R. Whitworth, J. R. Choate, S. J. Bissell, M. P. Moulton, and J. D. Hoffman. 2004. Geographic relationships of pocket gophers in southeastern Colorado. Trans. Nebr. Acad. Sci., 29:45-55.

Lessa, E. P. 1987. Functional morphology and allometry of the digging apparatus in picket gophers of the genera Thomomys, Geomys, and Cratogeomys (Rodentia: Geomyidae). Ph.D. Dissertation, New Mexico State Univ., Las Cruces, NM. 176 p.

-- and C. S. Thaeler, Jr. 1989. A reassessment of morphological specializations for digging in pocket gophers./. Mammal., 70:689-700.

MacArthur, R. H. 1972. Geographical Ecology: patterns in the distribution of species. Harper and Row Publishers, Inc., New York, NY.

Miller, R. S. 1964. Ecology and distribution of pocket gophers (Geomyidae) in Colorado. Ecology, 45:256-272.

--. 1967. Pattern and process in competition. Adv. Ecol. Res., 4:1-74.

Moulton, M. P., J. R. Choate, and S. J. Bissell. 1979. Sympatry of pocket gophers on Mesa de Maya, Colorado. Trans. Nebr. Acad. Sci., 82:194-195.

--, --, and-. 1983. Biogeographic relationships of pocket gophers in southeastern Colorado. Southwest. Nat., 28:53-60.

Neet, C. R. and J. Hausser. 1990. Habitat selection in zones of parapatric contact between the common shrew Sorex araneus and Millet's shrew S. coronatus. J. Anim. Ecol., 59:235-250.

Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecol. Model., 190:231-259.

-- and M. Dudik. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography, 31:161-175.

R Development Core Team. 2008. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org. Accessed 27 June 2013.

Riordan, E. C. and P. W. Rundel. 2009. Modelling the distribution of a threatened habitat: the California sage scrub. J. Biogeography, 36:2176-2188.

Segurado, P. and M. B. Araujo. 2004. An evaluation of methods for modelling species distributions. J. Biogeography, 31:1555-1568.

Sheppard, D. H. 1971. Competition between two chipmunk species (Eutamias). Ecology, 52:320-329.

Slade, N. A. and P. B. Robertson. 1977. Comments of competitively-induced disjunct allopatry. Occasional Papers of the Museum of Natural History, University of Kansas, 65:1-8.

Smolen, M. J., H. H. Genoways, and R. J. Baker. 1980. Demographic and reproductive parameters of the yellow-cheeked pocket gopher (Pappogeomys castanops). J. Mammal., 61:224-236.

States, J. B. 1976. Local adaptations in chipmunk (Eutamias amoenus) populations and evolutionary potential at species borders. Ecol. Monogr., 46:221-256.

Swets, J. A. 1988. Measuring the accuracy of diagnostic systems. Science, 240:1285-1293.

Terborgh, J. 1971. Distribution on environmental gradients: theory and a preliminary interpretation of distributional patterns in the avifauna of the Cordillera Vicabamba, Peru. Ecology, 52:32-40.

Thaeler, C. S. 1968. An analysis of the distribution of pocket gopher species in northeastern California (Genus Thomomys). U. Caif. Publ. Zool., 86:1-46.

Vaughan, T. A. 1963. Movements made by two species of pocket gopher. Am. Midi. Nat., 69:367-372.

--. 1967. Two parapatric species of pocket gophers. Evolution, 21:148-158.

Williams, L. R. and G. N. Cameron. 1984. Demography of dispersal in Attwater's pocket gopher, Geomys attwateri. J. Mammal., 67:216-224.

JUSTIN D. HOFFMAN (1)

Department of Biological and Health Sciences, McNeese State University, Lake Charles, Louisiana 70609

(1) Corresponding author: Telephone: 337.475.5659; e-mail: jhoffman@mcneese.edu

TABLE 1.--List of variables included in the predictive model
for each species. Also included are the numbers of presence
localities used (n), the training and testing area under the
curve (AUC) values, and omission rates using the minimum
training presence (MTP), and fixed 10% cumulative
probability thresholds. Asterisks indicate significance
level from a 1-tailed binomial probability (* = P < 0.05, **
= P < 0.001, *** = P < 0.0001)

Species             Model variables         n    Train AUC

Cratogeomys    % sand, % clay, land use,    44     0.85
castanops        permeability, porosity

Geomys         % sand, % clay, land use,    56     0.91
bursarius        rock volume

Species        Test AUC     MTP     10% fixed

Cratogeomys      0.89     0.00 **   0.00 ***
castanops

Geomys           0.81     0.06 *     0.2 **
bursarius
COPYRIGHT 2015 University of Notre Dame, Department of Biological Sciences
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2015 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Hoffman, Justin D.
Publication:The American Midland Naturalist
Article Type:Report
Geographic Code:1USA
Date:Apr 1, 2015
Words:5691
Previous Article:Interspecific interactions and habitat use by prairie voles (Microtus ochrogaster) and meadow voles (M. pennsylvanicus).
Next Article:Movements and habitat interactions of white-tailed deer: implications for chronic wasting disease management.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters