Habitat Use by Sonoran Desert Tortoises

Management and Conservation Article

Habitat Use by Sonoran Desert Tortoises

ERIN R. ZYLSTRA,1 School of Natural Resources, University of Arizona, 325 Biological Sciences East, Tucson, AZ 85721, USA ROBERT J. STEIDL, School of Natural Resources, University of Arizona, 325 Biological Sciences East, Tucson, AZ 85721, USA

ABSTRACT The distribution of desert tortoises (Gopherus agassizii) spans a wide range of biotic and abiotic conditions in the southwestern

United States and northwestern Mexico, with physical and behavioral differences distinguishing tortoises inhabiting the Mojave Desert from those inhabiting the Sonoran Desert. Relative to tortoise populations in the Mojave Desert, populations in the Sonoran Desert have not been well-studied. To assess how habitat use of desert tortoises in the Sonoran Desert was influenced by topography, vegetation, geomorphology, and soil, we surveyed 40 randomly located 3-ha sites for presence of adult tortoises within a site-occupancy framework. We modeled both occupancy and detection probability as a function of environmental features, and compared those results with a logistic regression model that assumed detection probability was equal to 1. Results from both approaches agreed, suggesting that habitat selection of tortoises in the Sonoran Desert was influenced primarily by topographic and geomorphologic features rather than by vegetation. Specifically, tortoises were more likely to occupy sites that were steep (we detected tortoises on 29% of sites with mean slope ,58 and 92% of sites with mean slope .158) and predominantly east-facing (53% of sites with ,5% of site facing E and 92% of sites with .20% facing E), and less likely to occupy northfacing slopes (100% of sites with ,10% of site facing N and 14% of sites with .60% facing N). Our results contrast with patterns of habitat use in the Mojave Desert where tortoises primarily occupy valley bottoms. Habitat use of tortoises in Sonoran and Mojave Desert populations differ considerably, contributing to the mounting body of evidence suggesting that these geographically distinct populations may represent separate species. (JOURNAL OF WILDLIFE MANAGEMENT 73(5):747?754; 2009)

DOI: 10.2193/2008-446

KEY WORDS desert tortoise, detection probability, Gopherus agassizii, habitat, site occupancy, Sonoran Desert.

Developing effective management strategies for populations of conservation concern requires a clear understanding of the environmental features that constitute habitat for the species (Noss et al. 1997). For species with geographic distributions that span a wide range of biotic and abiotic conditions, environmental features important for their conservation could vary markedly across their ranges. The geographic range of one such species, the desert tortoise (Gopherus agassizii), encompasses the Mojave and Sonoran Deserts of the United States and Mexico, and thornscrub and deciduous forests of western Sonora and northern Sinaloa (Germano et al. 1994, Van Devender 2002, Stebbins 2003). Tortoises north and west of the Colorado River, designated as Mojave desert tortoises by the United States Fish and Wildlife Service, have been studied extensively during the past 2 decades after disease and habitat loss resulted in severe population declines and their eventual listing as threatened in 1990 under the Endangered Species Act (U.S. Fish and Wildlife Service 1990, 1994; Christopher et al. 2003). Tortoise populations in the Sonoran Desert and tropical deciduous forest were not listed, however, in part because less is known about the ecology, habitat requirements, and status of populations outside of the Mojave Desert. Without a clear understanding of the habitat requirements for tortoises in these regions or long-term data necessary to reliably gauge population trends for a species with a 25-year generation time, the conservation outlook for tortoises in areas outside of the Mojave Desert is unclear.

Tortoises in the Mojave Desert inhabit valley bottoms with loamy soils, occasionally extending into washes along the lower bajada slopes and rocky hillsides (Bury et al. 1994,

1 E-mail: erinzylstra@

Germano et al. 1994, Andersen et al. 2000). In contrast, tortoises in the Sonoran Desert inhabit rocky hillsides, mountain foothills, and incised washes and are thought to be rare in valley bottoms (Barrett 1990, Germano et al. 1994, Bailey et al. 1995, Riedle et al. 2008). Previous descriptions of habitat used by tortoises in the Sonoran Desert have not accounted for potential biases introduced by variation in detection probability that can result in areas being classified incorrectly as unused (Rettie and McLoughlin 1999).

Besides differences in habitat use, tortoises in the Mojave and Sonoran regions differ morphologically (Weinstein and Berry 1987, Germano 1993), genetically (Lamb et al. 1989, Murphy et al. 2007), physiologically (Turner et al. 1986, Wallis et al. 1999, Averill-Murray 2002), and behaviorally (Woodbury and Hardy 1948, Duda et al. 1999, AverillMurray et al. 2002). These marked differences have led experts to suggest that desert tortoise populations in the Mojave and Sonoran Deserts might represent distinct species (Berry et al. 2002, Murphy et al. 2007).

Given the question of whether Mojave and Sonoran populations represent separate species and the lack of reliable, quantitative assessments of habitat for Sonoran desert tortoises, evaluating patterns of habitat use by tortoises in the Sonoran Desert and contrasting those with habitat used by tortoises in the Mojave Desert seems timely. Consequently, our primary objective was to assess habitat use of tortoises in the Arizona Upland subdivision of the Sonoran Desert within the framework of site occupancy (MacKenzie et al. 2002), which would allow us to account for potential biases associated with imperfect detection (Gu and Swihart 2004, Mazerolle et al. 2005, MacKenzie 2006). Additionally, we sought to compare results from site occupancy models to results from a logistic regression

Zylstra and Steidl Habitat Use by Sonoran Desert Tortoises

747

Figure 1. Study areas and locations of occupancy sites for Sonoran desert tortoises in the Rincon Mountains in 2005 and Tucson Mountains in 2006 in Saguaro National Park, southern Arizona, USA.

model, which assumes that detection probability is equal to 1.

STUDY AREA

We studied tortoises in the eastern portion of the Sonoran Desert in southern Arizona, USA, a region with high tree and shrub diversity and limited annual rainfall of 240?300 mm that falls in a seasonally bimodal pattern (Turner and Brown 1982). We studied 2 areas of Saguaro National Park, one in 2005 and one in 2006 (Fig. 1). In 2005, we studied a 43-km2 area in the Rincon Mountains that ranged in elevation from 800 m to 1,150 m and was characterized by Arizona Upland vegetation including foothills paloverde (Parkinsonia microphylla), saguaro (Carnegiea gigantea), velvet mesquite (Prosopis velutina), acacia (Acacia spp.), and cholla (Opuntia spp.; Turner and Brown 1982, Bowers and McLaughlin 1987). The Rincon Mountain study area encompassed 2 distinct geomorphologic regions. One-third of the study area was flat with broad, sandy washes and few rock or boulder formations, and the remaining two-thirds were topographically variable with numerous rock formations. In 2006, we studied an 80-km2 area in the Tucson Mountains that ranged in elevation from 675 m to 1,150 m and was located on the edge of a transition zone between the Arizona Upland and Lower Colorado River Valley subdivisions of the Sonoran Desert in Avra Valley, Arizona (Turner and Brown 1982). Creosote bush (Larrea tridentata)

and bursage (Ambrosia spp.), often associated with the Lower Colorado River Valley subdivision, were abundant in the northwestern region of the mountains (Rondeau et al. 1996). Topography in this area varied from steep slopes with jagged volcanic rocks to lowland flats. Foothills in the Tucson Mountains contained numerous washes with incised banks and caliche caves, which desert tortoises often used for shelter.

METHODS

Field Surveys Within each study area, we randomly located 20 3-ha sites (170 m 3 170 m). In the Rincon Mountains, we allocated sites in proportion to areal coverage of the 2 geomorphologic regions. In the Tucson Mountains we selected sites completely at random, after excluding areas .500 m from hills from the sampling frame because we wanted to assess habitat use in areas where the density of tortoises was .0.05 tortoises/ha (Averill-Murray and Averill-Murray 2005). In each study area we excluded from the sample sites that crossed roads, were too steep to access or survey safely, or were ,170 m from another site.

We used repeated presence?absence surveys to estimate the proportion of sites occupied by tortoises (MacKenzie et al. 2002). We surveyed each site 5 times between early July and mid-October because tortoises in the Sonoran Desert are most active during the summer monsoon season (Van Devender 2002); repeated surveys of the same site were separated by !7 days. We surveyed sites during mornings (0545?1200 hr) and evenings (1600?2000 hr), with each site surveyed at least twice in the morning and twice in the evening. One to 5 observers surveyed each site by walking parallel lines spaced approximately 10 m apart, while scanning open ground, looking under vegetation, and using mirrors or flashlights to inspect holes and crevices. During the first survey of each site in 2005, searches ended when we found an adult tortoise (midline carapace length !180 mm; Germano 1994), and for surveys 2?5 in 2005 and all surveys in 2006, we searched each site in its entirety. We classified a site as occupied if we observed !1 adult tortoise on the site during any of the 5 surveys. We did not include observations of tortoises with carapace length ,180 mm in estimates of occupancy. In our study and in previous studies, only small proportions of overall detections were juvenile or subadult tortoises (Swann et al. 2002, Averill-Murray and AverillMurray 2005), probably because detection probabilities for these age classes are lower than for adults (Berry and Turner 1986, Morafka 1994, Anderson et al. 2001).

Environmental Features We evaluated the influence of 9 topographic, vegetative, geologic, and soil features on habitat selection of tortoises (Table 1), each of which has been suggested to affect distribution and habitat use of desert tortoises in different parts of their range (Woodbury and Hardy 1948, Andersen et al. 2000, Riedle et al. 2008). We derived aspect, slope, and elevation from a 1:24,000-scale digital elevation model, and for elevation and slope, we used spatial analysis tools in

748

The Journal of Wildlife Management 73(5)

Table 1. Environmental and survey features evaluated as covariates for occupancy and detection probability of Sonoran desert tortoises in southern Arizona, USA, 2005?2006. Phrases in parentheses are labels used to represent features in models in subsequent tables.

Feature

Variables

Description

Yr Elevation Slope Aspect

Geologic class (GeoClass)

Vegetation community (VegCom)

Soil class (SoilClass)

Plant cover (Cover) Incised wash (Wash) Shelter-sites (Shelter) Observer experience (Obs) Survey period (t)

Year ElevMean SlopeMean AspN AspE AspS GeoRD GeoSed GeoIgneo VegMixed VegPami VegPrve SoilPsc

SoilApc SoilDl Cover Wash Shelter Obs

t

Indicator variable for 2005 Standardized mean elevation of site (m) Standardized mean slope of site (8) Proportion of site N-facing Proportion of site E-facing Proportion of site S-facing Recent sedimentary deposits present on !20% of site Sedimentary rock present on !20% of site Igneous rock present on !20% of site Mixed shrub and cacti vegetation community present on !20% of site Foothills-paloverde?dominated vegetation community present on !20% of site Velvet-mesquite?dominated vegetation community present on !20% of site Pinaleno?Stagecoach?Cave or Glendale?Arizo?Hantz soil types present on

!20% of site Anklam?Pantano?Cellar soil type present on !20% of site Deloro?Lampshire?Rock Outcrop soil type present on !20% of site Proportion of site with tree and dense shrub cover Indicator variable for presence of !1 incised wash No. of potential tortoise shelter-sites on each site Proportion of observers with experience searching for Sonoran desert

tortoises on each survey Separate estimates of detection probability for each survey period

ArcView 3.3 to calculate mean values for each site that we standardized ([raw score ? mean score]/SD). For aspect, we generated 4 variables to represent the proportion of each site with slopes facing each of the cardinal directions; we used 3 of 4 variables for analyses because the fourth variable was almost perfectly predicted by the other three.

We derived soil and geologic classifications from digital soil and geology maps and classified vegetation communities in the field. We distinguished 3 general soil classes: Pinaleno?Stagecoach?Cave or Glendale?Arizo?Hantz, Anklam?Pantano?Cellar, and Deloro?Lampshire?Rock (U.S. Department of Agriculture 2003); 3 geologic classes: recent sedimentary deposits, sedimentary rock, and igneous rock; and 3 vegetation communities: foothills-paloverde?dominated, velvet-mesquite?dominated, and mixed-shrub and cacti, which were occasionally dominated by jojoba (Simmondsia chinensis). To characterize soil, geology, and vegetation communities, we created binary indicator variables representing whether a particular soil class, geologic class, or vegetation community covered !20% of a site.

We estimated plant cover, presence of incised washes, and the number of potential tortoise shelter-sites during field observations and with aerial photographs. To estimate the proportion of each site covered by trees and shrubs, we overlaid a dot grid with 10-m spacing on color ortho-photos (Nowak et al. 1996). Given the structure of woody vegetation in this region, it was impossible to distinguish between trees and certain shrubs; therefore, we report cover for all trees and dense shrubs combined, including desert hackberry (Celtis pallida), jojoba, and acacia. We used field observations and aerial photographs to determine whether each site had !1 incised wash (sides of the wash extending !1.5 m above the bottom of the wash). Finally, in the field

we estimated the number of potential tortoise shelter-sites on each site, with shelter-sites defined as rock dens or soil burrows .0.5 m deep whose lateral dimensions did not greatly exceed that of an adult tortoise (ht or width ,0.5 m). We used logistic regression for binomial counts to describe the relationship between number of shelter-sites and the proportion of surveys with detections of adult tortoises on a site.

Occupancy To determine which environmental features affected site occupancy of tortoises, we modeled occupancy (w) and detection probability (p) in Program PRESENCE 2.1 in 3 steps (MacKenzie et al. 2002). We first determined which environmental features explained variation in w while maintaining a general model for p and then determined which environmental or survey features explained variation in p while maintaining a general model for w. Finally, we combined results from the first 2 steps to identify features to include in our set of candidate models. In the first step, we created a general model for p based on a set of potentially influential variables: year, survey period, proportion of observers on each survey with experience searching for Sonoran desert tortoises, and proportion of each site covered by trees and dense shrubs (Table 1). Because we surveyed different areas in 2005 and 2006, effects of year and area are confounded; we refer to this term in our analyses as year. We then divided environmental features with the potential to influence w into 3 groups: topography (i.e., aspect, elevation, and slope), geomorphology (i.e., geology, soil, and incised wash), and vegetation (i.e., vegetation community and tree and shrub cover). Within each group, we created models for w with every combination of environmental

Zylstra and Steidl Habitat Use by Sonoran Desert Tortoises

749

features while maintaining the same general model for p (MacKenzie 2006). In this and all other analyses, we ranked models within each group according to Akaike's Information Criterion corrected for small sample sizes (AICc), and considered top models to be those with DAICc 2.0, which we included in subsequent analyses (Burnham and Anderson 2002). In the second step, we created a general model for w including all environmental features from the top models identified in the first step, plus a covariate for year. We then created models for p with every combination of features thought to potentially influence p (yr, survey period, proportion of observers with experience, and proportion of each site covered by trees and dense shrubs) while maintaining the general model for w. In the third step, we created a set of candidate models that included all combinations of environmental and survey features included in top models for w and p and assessed goodness-of-fit for the most general candidate model (MacKenzie and Bailey 2004). We retained models with DAICc 2.0 for inference, and we adjusted AICc model weights so that weights of the top-ranking models summed to 100% (MacKenzie and Bailey 2004).

For each top-ranking model, we generated an overall estimate of occupancy by averaging site-specific estimates. To include components of variation that reflected sampling error associated with covariates, we used the delta method (Seber 1982, MacKenzie et al. 2006, Powell 2007) to construct a variance?covariance matrix for site-specific occupancy estimates, which we used to calculate the standard error of the overall occupancy estimate (MacKenzie et al. 2006, Zylstra 2008). Similarly, we used survey-specific estimates of detection probability to generate an estimate of detection probability and associated standard error for each top-ranking model and then generated model-averaged estimates across all top-ranking models.

Logistic Regression We used logistic regression to identify habitat features that affected site occupancy under the assumption that detection probability was equal to 1. We considered a site unoccupied if we failed to detect an adult tortoise during all 5 surveys and occupied if we detected an adult tortoise during any survey. Similar to occupancy analyses, we began by evaluating environmental features in topography, geomorphology, and vegetation groups. Within each group, we considered a feature to have potential explanatory power if !1 of the variables used to describe the feature resulted in a likelihood-ratio test with P , 0.20. We then included features with potential explanatory power in a full model. We used drop-in-deviance tests to compare the full model with a series of reduced models, each of which excluded the set of variables used to describe one environmental feature (Ramsey and Schafer 2002). If the drop-in-deviance test yielded P , 0.05, we concluded that the environmental feature excluded from the reduced model affected occupancy. We then included all these features in a final model used to explain variation in occupancy.

Table 2. Results of models examining relationships between environmental features and probability of occupancy by Sonoran desert tortoises in southern Arizona, USA, 2005?2006. Results are based on a general model for detection probability: p(Cover?Obs?Year?t).

Models

Ka log Lb AICcc DAICcd wie

Topography group

w(Slope)

10 ?103.67 234.93 0.00 0.41

w(Elevation?Slope)

11 ?102.10 235.62 0.69 0.29

w(Aspect?Slope)

13 ?97.97 235.94 1.01 0.25

w(Aspect?Elevation?Slope) 14 ?97.63 240.07 5.14 0.03

w(Aspect)

12 ?102.72 240.99 6.06 0.02

w(Aspect?Elevation)

13 ?101.55 243.09 8.16 0.01

w(.)

9 ?110.76 245.51 10.58 0.00

w(Elevation)

10 ?110.20 247.99 13.06 0.00

Geomorphology group

w(.)

9 ?110.76 245.51 0.00 0.63

w(Wash)

10 ?109.56 246.71 1.20 0.34

w(GeoClass)

12 ?108.70 252.95 7.44 0.02

w(GeoClass?Wash)

13 ?107.57 255.14 9.63 0.01

w(SoilClass?Wash)

13 ?107.77 255.54 10.03 0.00

w(SoilClass)

12 ?110.08 255.72 10.21 0.00

w(GeoClass?SoilClass)

15 ?108.37 266.74 21.23 0.00

w(GeoClass?SoilClass

?Wash)

16 ?106.67 268.99 23.48 0.00

Vegetation group

w(.)

9 ?110.76 245.51 0.00 0.83

w(Cover)

10 ?110.58 248.74 3.23 0.16

w(VegCom)

12 ?109.71 254.98 9.47 0.01

w(Cover?VegCom)

13 ?109.65 259.29 13.78 0.00

a K ? no. of parameters. b log L ? log-likelihood. c AICc ? Akaike's Information Criterion corrected for small sample sizes. d DAICc ? AICc relative to the most parsimonious model. e wi ? AICc model wt.

RESULTS

We surveyed 40 sites (20 in 2005, 20 in 2006) 5 times each between 7 July and 14 October, during which we observed adult tortoises on 27 sites (67.5%) and during 61 of 200 surveys (30.5%). We observed 0?3 adults per survey with an average of 0.38 adults (SE ? 0.05) per survey. We observed 76 adult tortoises (!61 unique individuals) and 23 juvenile or subadult tortoises (20 unique individuals).

Number of potential shelter-sites per site was a strong predictor of the proportion of surveys on which we detected adult tortoises (v21 ? 41.88, P 0.001). Specifically, for each additional shelter-site on a site, odds of detecting a tortoise increased by a factor of 1.32 (95% CI ? 1.21?1.45). Models that included number of shelter-sites as a covariate for w, however, yielded unrealistic standard errors because estimates approached parameter boundary values (w ? 0 or 1; Anderson 2008). Despite the clear importance of this local feature, we were forced to exclude it from analyses of the broader scale environmental features we evaluated.

When considering only topographic variables, slope was the most important feature affecting occupancy and accounted for 97% of model weights (Table 2). When considering only geomorphic variables, presence of incised washes explained more variation in occupancy than geologic or soil class and accounted for 35% of model weights versus only 2% for geologic variables and 1% for soil variables. When considering only vegetation community and plant

750

The Journal of Wildlife Management 73(5)

Table 3. Results of models examining relationships between environmental and survey features and detection probability of Sonoran desert tortoises in southern Arizona, USA, 2005?2006. Results are based on a general model for occupancy: w(Aspect?Elevation?Slope?Wash?Year).

Models

Ka

log Lb

AICcc DAICcd wie

p(.)

9 ?103.24 230.48 0.00 0.28

p(Year)

10 ?101.89 231.36 0.88 0.18

p(Obs)

10 ?102.32 232.22 1.74 0.12

p(Obs?Year)

11 ?100.45 232.32 1.84 0.11

p(Cover?Year)

11 ?100.63 232.69 2.21 0.09

p(Cover)

10 ?102.66 232.90 2.42 0.08

p(t)

13 ?96.59 233.17 2.69 0.07

p(Cover?Obs?Year)

12 ?99.19 233.93

3.45 0.05

p(Obs?Cover)

11 ?101.81 235.03 4.55 0.03

p(Year?t)

14 ?99.66 244.11 13.63 0.00

p(Cover?t)

14 ?100.46 245.72 15.24 0.00

p(Obs?t)

14 ?100.51 245.81 15.33 0.00

p(Cover?Year?t)

15 ?98.36 246.72 16.24 0.00

p(Obs?Year?t)

15 ?98.67 247.33 16.85 0.00

p(Cover?Obs?t)

15 ?99.96 249.92 19.44 0.00

p(Cover?Obs?Year?t) 16 ?97.37 250.38 19.90 0.00

a K ? no. of parameters. b log L ? log-likelihood. c AICc ? Akaike's Information Criterion corrected for small sample sizes. d DAICc ? AICc relative to the most parsimonious model. e wi ? AICc model wt.

cover, neither feature was strongly associated with occupancy. Because all models with vegetation features had considerably higher AICc values than a model with no covariates for occupancy, no vegetation features were retained as covariates in subsequent analyses. Finally, year and observer experience best explained variation in detection probability, accounting for 43% and 30% of total model weights, respectively (Table 3). There was little evidence that detection probability varied with plant cover or survey period.

We considered 11 of 63 candidate models examined to be top-ranking models that accounted for 66% of total model weights; we excluded one model because estimates did not converge (Table 4). Goodness-of-fit for the most general model in the candidate set was acceptable (v21 ? 5.59, P ?

0.78). We estimated occupancy to be 0.71 (95% CI ? 0.59? 0.83) and detection probability to be 0.43 (95% CI ? 0.34? 0.53) across top-ranking models. Occupancy increased as the mean slope of a site increased, and all sites with mean slope !168 were occupied (Fig. 2). Aspect was also an important predictor of occupancy (Table 4), which was positively associated with east-facing slopes and negatively associated with north- and south-facing slopes (Fig. 2). Elevation and presence of incised washes were not strong predictors of occupancy, with each feature having summed model weights across all candidate models of ,35%.

Detection probabilities increased somewhat with observer experience and varied between years, although models with constant detection probability often ranked higher than models that included covariates. Mean detection probability was 0.24 (95% CI ? 0.09?0.50, from w[Slope] p[Observer]) when ,50% of observers had previous experience, 0.32 (95% CI ? 0.20?0.48) when 50?75% of observers had previous experience, and 0.44 (95% CI ? 0.35?0.54) when all observers had previous experience. Detection probability was also higher in 2005 than 2006 (2005: p ? 0.52, 95% CI ? 0.39?0.65; 2006: p ? 0.38, 95% CI ? 0.28?0.50, from w[Aspect?Slope] p[Year]).

For the logistic regression analysis that assumes detection probability is equal to 1, occupancy was best predicted by slope and aspect (slope: v21 ? 9.31, P ? 0.002; aspect: v23 ? 11.37, P ? 0.010). Occupancy was positively associated with mean slope of the site and proportion of the site with eastfacing slopes and negatively associated with north- and south-facing slopes (Table 5).

DISCUSSION

Sonoran and Mojave Populations Our finding that tortoises in the Sonoran Desert were more likely to inhabit steep slopes contrasts sharply with tortoises in the Mojave Desert, which studies suggest inhabit valley bottoms and other flat areas primarily (Germano et al. 1994). Although tortoises in the Sonoran Desert occasionally inhabit intermountain valleys and tortoises in the

Table 4. Top-ranking models describing occupancy of Sonoran desert tortoises in southern Arizona, USA, 2005?2006.

Models

Ka

log Lb

AICcc

wid

w^ e

w(Aspect?Slope) p(.)

6

?103.91

222.35

0.14

0.69

w(Aspect?Slope) p(Year)

7

?102.66

222.81

0.11

0.68

w(Slope) p(.)

3

?108.12

222.90

0.11

0.74

w(Slope) p(Obs)

4

?106.95

223.05

0.10

0.74

w(Aspect?Slope) p(Obs?Year)

8

?101.22

223.09

0.10

0.68

w(Slope?Wash) p(Obs)

5

?105.67

223.10

0.10

0.78

w(Aspect?Slope) p(Obs)

7

?102.96

223.43

0.08

0.69

w(Elevation?Slope) p(.)

4

?107.27

223.69

0.07

0.71

w(Elevation?Slope) p(Obs)

5

?106.17

224.11

0.06

0.71

w(Elevation?Slope) p(Obs?Year)

6

?104.80

224.15

0.06

0.70

w(Slope) p(Obs?Year)

5

?106.25

224.25

0.06

0.71

a K ? no. of parameters. b log L ? log-likelihood. c AICc ? Akaike's Information Criterion corrected for small sample sizes. d wi ? AICc model wt. e w^ ? overall estimate of occupancy. f SE(w^ ) ? SE of overall occupancy estimate.

SE(w^ )f

0.05 0.05 0.05 0.05 0.05 0.03 0.05 0.06 0.06 0.06 0.08

Zylstra and Steidl Habitat Use by Sonoran Desert Tortoises

751

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download