Journal of Mammalogy, 91(1):66–78, 2010
Landscape effects on diets of two canids in northwestern Texas: a multinomial modeling approach PATRICK R. LEMONS,* JAMES S. SEDINGER, MARK P. HERZOG, PHILLIP S. GIPSON,
AND
RICK L. GILLILAND
Department of Natural Resources and Environmental Sciences, University of Nevada, Reno, Reno, NV 89512, USA (PRL, JSS, MPH) United States Geological Survey, Kansas Cooperative Fish and Wildlife Research Unit, Division of Biology, Kansas State University, 205 Leasure Hall, Manhattan, KS 66506-3501, USA (PSG) United States Department of Agriculture–Animal and Plant Health Inspection Service–Wildlife Service, Canyon District, P.O. Box 60277, West Texas A&M University, Canyon, TX 79016, USA (RLG) Present address of PRL: United States Geological Survey–Alaska Science Center, 4210 University Drive, Anchorage, AK 99508, USA Present address of MPH: Point Reyes Bird Observatory Conservation Science, 4990 Shoreline Highway, Stinson Beach, CA 94924, USA * Correspondent:
[email protected] Analyses of feces, stomach contents, and regurgitated pellets are common techniques for assessing diets of vertebrates and typically contain more than 1 food item per sampling unit. When analyzed, these individual food items have traditionally been treated as independent, which represents pseudoreplication. When food types are recorded as present or absent, these samples can be treated as multinomial vectors of food items, with each vector representing 1 realization of a possible diet. We suggest such data have a similar structure to capture histories for closed-capture, capture–mark–recapture data. To assess the effects of landscapes and presence of a potential competitor, we used closed-capture models implemented in program MARK into analyze diet data generated from feces of swift foxes (Vulpes velox) and coyotes (Canis latrans) in northwestern Texas. The best models of diet contained season and location for both swift foxes and coyotes, but year accounted for less variation, suggesting that landscape type is an important predictor of diets of both species. Models containing the effect of coyote reduction were not competitive (DQAICc 5 53.6685), consistent with the hypothesis that presence of coyotes did not influence diet of swift foxes. Our findings suggest that landscape type may have important influences on diets of both species. We believe that multinomial models represent an effective approach to assess hypotheses when diet studies have a data structure similar to ours. DOI: 10.1644/07-MAMM-A-291R1.1. Key words: Canis latrans, capture–mark–recapture, coyotes, diets, multinomial data, multiple responses, pellet analysis, scat analysis, stomach analysis, Vulpes velox
E 2010 American Society of Mammalogists
concern for its conservation remain (Schauster et al. 2002; Scott-Brown et al. 1987). Numerous studies have documented the direct impact of coyotes on populations of swift foxes. The largest documented cause of mortality of swift foxes is depredation by coyotes (Carbyn et al. 1994; Covell 1992; Kamler et al. 2003; Matlack et al. 2000; Sovada et al. 1998). However, killing of swift foxes by coyotes was not typical predation; swift foxes were killed and buried, but coyotes did not appear to consume them (Kitchen et al. 1999; Matlack et al. 2000). Several studies have
Historically, swift foxes (Vulpes velox) were abundant across the shortgrass and midgrass prairies from southern Canada to west-central Texas and eastern New Mexico (Egoscue 1979). Beginning in the mid-19th century, numbers of swift foxes declined rapidly with expansion of human settlements, habitat destruction, and poisoning used to control depredation of livestock by wolves (Canis lupus) and coyotes (Canis latrans—Egoscue 1979; Hines 1980; Scott-Brown et al. 1987). Populations of swift foxes may have increased slightly in the mid-20th century as use of poisons was reduced, but populations remain depressed (Egoscue 1979; Kilgore 1969; Samuel and Nelson 1992; Scott-Brown et al. 1987). Questions about the ecology of the swift fox and a general
www.mammalogy.org 66
February 2010
LEMONS ET AL.—VARIATION IN CANID DIETS
demonstrated that removal of coyotes results in changes in the local faunal community (Henke 1992; Henke and Bryant 1999; Linhart and Robinson 1972). Change in diets of swift foxes following removal of coyotes also suggests that coyotes have indirect effects on the foraging behavior and diets of swift foxes (Kitchen et al. 1999). Numbers of swift foxes have declined in association with conversion of shortgrass and midgrass prairies to agriculture, leading some to suggest that swift foxes may be habitat specialists (Egoscue 1979; Kamler et al. 2003; Matlack et al. 2000; Scott-Brown et al. 1987). However, a lack of understanding exists about the effects of habitat on populations of swift foxes. Habitat did not affect survival or diets of swift foxes in Kansas (Sovada et al. 1998, 2001). However, Matlack et al. (2000) found that swift foxes in habitats consisting of dryland agriculture had lower body condition and higher mortality than those in grassland habitats. Furthermore, Kamler et al. (2003) found that swift foxes in northwestern Texas completely avoided irrigated agricultural habitats and used dryland agricultural habitats and conservation reserve fields substantially less than expected. Taken together, these studies suggest that swift foxes may be grassland habitat specialists (Kamler et al. 2003). Because of the potential relationship between habitat and population dynamics of swift foxes, we were interested in the relationship between habitat type and diets of swift foxes. Landscape type and presence of a potential competitor may both affect diets of swift foxes, which may in turn affect body condition and survival. The focus of this study was to determine whether 1 or both factors were important predictors of the diets and body conditions of swift foxes. The most commonly used technique for assessing diets of many vertebrates has been analyses of different types of feces, regurgitated pellets, or stomach contents (Litvaitis et al. 1996) using contingency tables (Azevedo et al. 2006; Glen and Dickman 2006; Phillips et al. 2007), analysis of variance (ANOVA—Badzinski and Petrie 2006; Zeppelin and Ream 2006), or similar techniques. Numerous studies have documented problems with estimating dietary intake from these types of samples, however, and have offered methods for making estimation more accurate (Floyd et al. 1978; Kelly and Garton 1997; Lockie 1959; Spaulding et al. 2000; Weaver and Hoffman 1979). We are unaware of any study addressing how dietary information such as these should be analyzed quantitatively. We use multinomial models developed for analyzing capture–mark–recapture data from closed populations to assess hypotheses about the diets of swift foxes and to estimate contributions of particular food items to the diet. Similar analyses have been used to model and estimate occupancy rates using presence–absence data (MacKenzie et al. 2002, 2006), which is analogous to presence or absence of food items in diet samples. No previous study has assessed diets among vertebrates by this method, which therefore requires background and justification. In most diet studies, sampling units (i.e., feces, stomach samples, or pellets—Sokal and Rohlf 1999; Zar 1999)
67
typically contain more than 1 food item (Bonesi et al. 2004; Kamler et al. 2007a, 2007b; Phillips et al. 2007). Traditional analyses assume that multiple food items within the sampling unit are independent of each other. Individual variation in habitat use, dietary preferences, or spatial or temporal correlation in the availability of food items makes it unlikely that individual food items within a sample are truly independent (Thomas and Taylor 1990). Furthermore, each sampling unit (i.e., scats, stomachs, or pellets) now appears in the analysis more than once, accounting for the multiple food items within the sampling unit. Failure to account for dependence among food items violates the assumption of independent responses and therefore represents pseudoreplication (Hurlbert 1984). We envision 2 general levels of diet data. First, if food items are quantitatively measured, as percent by weight or percent volume for each sample, then each sample can be treated as a multivariate vector whose dimension is defined by the number of types of food items in the study. Multivariate statistical procedures would be appropriate for such data. Second, data from diet studies often are in the form of presence–absence data (1 for presence of a particular food type and 0 if absent). Such data can take the form of a series of 1s (items present) and 0s (items absent) from each sampling unit, producing a vector of the form ‘‘01100010’’ for a situation where there were 8 distinct food categories available in the sampling unit. This data form is identical to that for closed-capture capture–mark–recapture studies, which can be analyzed using maximum-likelihood approaches for multinomial data (Otis et al. 1978). Each individual in a capture– mark–recapture study is treated as a unique sampling unit with its own capture history consisting of 1s and 0s for encounter and nonencounter, respectively. In a capture–mark–recapture study an encounter probability for each occasion is estimated (Williams et al. 2002). For dietary data the encounter probability would be interpreted as the proportion of individuals containing a particular food item. Our study to determine how landscape type and presence of a potential competitor influence diets of swift foxes was completed in northwestern Texas. We collected feces (i.e., scats) of swift foxes and coyotes from January 1999 to December 2000 during all seasons and across 2 landscape types in the region. Coyote numbers were reduced on 1 site in 1 year, allowing us to assess the impact of coyotes on diets of swift foxes and coyotes. We use model selection techniques (Burnham and Anderson 2002) to address hypotheses about how landscape type and coyote density affect diets of swift foxes. In particular, we addressed the hypothesis that diets of swift foxes varied between habitats and in response to presence of coyotes.
MATERIALS AND METHODS Study site.—We conducted research at two 93-km2 study sites in northwestern Texas. The 1st study site was designated our treatment site and was located 55 km west of Stratford in
68
JOURNAL OF MAMMALOGY
Dallam County, Texas, predominantly within the Rita Blanca National Grasslands. The area was restored shortgrass prairie habitat dominated by blue grama (Bouteloua gracilis) and buffalograss (Buchloe dactyloides) and was moderately to heavily grazed by cattle. The 2nd study site was designated our control site and was located on a private ranch in Sherman County, Texas, approximately 12 km south of Stratford, Texas. This site was composed of shortgrass prairie rangeland (35%), cultivated fields (31%), and Conservation Reserve Program (35%). Crops included corn, winter wheat, and sorghum. Conservation Reserve Program land had either recently been enrolled in the program and was planted to warm-season grasses including sideoats grama (Bouteloua curtipendula), blue grama, sand dropseed (Sporobolus cryptandrus), and buffalograss, or had not been recently reenrolled in the program and was vegetated by old world bluestem (Bothriochloa spp.). Coyote reduction.—During 1999 coyote numbers were allowed to fluctuate under prestudy conditions. Beginning in January 2000 the United States Department of Agriculture– Animal Plant Health Inspection Service–Wildlife Services reduced coyote numbers by aerial hunting out of a Piper Super Cub aircraft. The removal effort was part of a concurrent study (Kamler et al. 2003) and provided us with the opportunity to document the impact of coyote abundance on diets of swift foxes. The Rita Blanca National Grasslands site was flown 4 times in year 2000 at approximately 3-month intervals (i.e., winter: January–March; spring: April–June; summer: July– September; and fall: October–December). During each of the 4 periods, coyote removal by shooting occurred for approximately 12 h over a 3- to 4-day period. Coyotes also were removed from a buffer zone approximately 2 km in width around this study site. A total of 227 coyotes were removed from this area (winter 5 97, spring 5 54, summer 5 13, fall 5 63), which resulted in a 56% decrease in relative abundance of coyotes from pre- to posttreatment (Kamler et al. 2003). Scat collection and preparation.—We collected scats during each of the 4 seasons from January 1999 through December 2000. Four 2-km transects per study site were walked once each season. Transects within study areas were separated on average by 2.5 km covering the majority of the study areas, and all were sampled within 2 days. A minimum of 20 scats from swift foxes and coyotes were collected during each season at both study sites. We also collected scats at den sites and opportunistically in some seasons to achieve a minimum sample size of 20 scats per site per season. Scats were identified to species according to size and shape (Murie 1974) and were individually bagged and labeled with date, season, and location. Following collection we placed scats in nylon mesh bags and soaked them in warm water with detergent for 30 min in an automatic washing machine. Following soaking, bags were washed for 2 cycles and allowed to air dry. Once bags dried we opened and identified the food remains. Reference collections and keys for teeth and hair were used to identify
Vol. 91, No. 1
prey remains (Gilbert 1990; Moore et al. 1997). To minimize bias associated with overestimation of food items in diets of carnivore, trace items (i.e., items comprising , 5% of individual scats) were excluded from analysis (Kelly 1991; Knowlton 1964). Body size.—Swift foxes were captured during fall and winter 1998, 1999, and 2000 using Havahart cage traps (Woodstream Corp., Lititz, Pennsylvania). The following information was collected from each individual: body mass, body length, tail length, sex, reproductive condition, age class, and injury type and severity, if any. Foxes were aged based on tooth wear and classified as adult (1 year old) or juvenile (,1 year old—Gier 1968). Our research protocol was approved by the Texas Tech University Institutional Animal Care and Use Committee. Statistical analysis.—Food items were identified and separated into 1 of 8 categories: insects, small rodents (average body mass , 100 g), large rodents (average body mass . 100 g), lagomorphs, ungulates, birds, crops, and vegetation. Thus, we characterized each scat by a vector of 8 different 1s and 0s, representing presence or absence, respectively, of food items from each category. For example, the following multinomial sequence, 11000001, would indicate that a fecal sample contained insects, small rodents, and vegetation but not large rodents, lagomorphs, ungulates, birds, or crops. In a closed-capture capture–mark–recapture study data are assumed to be multinomial; for each encounter occasion probability of encounter is estimated using maximum likelihood (Williams et al. 2002). We also assumed a multinomial data structure and estimated for each food type the probability it was present in the average individual. Our analyses required the following assumptions: scats represented independent samples of the diet, food categories were equally available to all individuals within class (e.g., swift foxes in the control site in spring 1999) analogous to equal encounter probability among individuals in a standard closed-capture analysis, and dietary items within samples were identified correctly. We analyzed diet data using Huggins’ (1989) models (equation 1) for closed populations within program MARK (White and Burnham 1999). We used Huggins’ (1989) models because such models condition population size out of the likelihood and population size was not meaningful for our data structure. The full likelihood was: ð1{xij Þ xij K 8 pij 1{pij , L~ P P ð1Þ i~1 j~1 1{ PK l~1 1{pil where pij was the probability individual i consumed food type j. The data structure consisted of j 5 1, …, 8 food types of which could have been consumed by individuals i, …, K. Each food type had to have been consumed at least once to be included. For each capture occasion, xij 5 1 if i is encountered and 0 if not encountered on occasion j (Huggins 1989). We then constrained initial encounter probability (pi) to equal reencounter probability (ci). Traditional closed-capture capture–mark–recapture analyses allow differences between
February 2010
LEMONS ET AL.—VARIATION IN CANID DIETS
initial encounter and reencounter probabilities representing behavioral responses to capture, which was not meaningful in our analyses. To control for correlation among food items we calculated the overdispersion parameter cˆ using a goodness-offit statistic as recommended by Anderson et al. (1994), Burnham and Anderson (2002), and Williams et al. (2002). Use of the parameter cˆ is conservative, and this conservative nature increases with increasing cˆ-values (Anderson et al. 1994). Therefore, the use of this parameter in our analysis should result in conservative estimates of model performance, and precision of parameter estimates. We calculated cˆ as 3.1815 using a goodness-of-fit statistic. Therefore, we used quasi-AICc (QAICc) values throughout our analysis. We built several models to determine which variables best described diet of swift foxes and coyotes. We included variables species, year, season, and landscape type in our models. Because coyote removal occurred in 1 landscape type only during the 2nd field season, we examined the hypothesis of coyote effects on diets of swift foxes using the performance of a model containing an interaction among species, year, and landscape. That is, if abundance or density of coyotes affected the diets of swift foxes, we expected diets of foxes to change only in the landscape where coyotes were removed relative to diets of swift foxes in the same landscape the previous year and in other landscapes in the same year. We used the performance of the models containing either an interactive or additive effect of landscape to assess the hypothesis that landscapes affected diet in either or both species. We used the same data set as that of Kamler et al. (2007a, 2007b) in a combined analysis to determine what factors were important for the diets of swift foxes and coyotes. Kamler et al. (2007a, 2007b) used a contingency table approach to determine factors affecting diets of swift foxes and coyotes. This approach, as discussed above, has serious problems with pseudoreplication. Therefore, we compared our results using a capture–mark– recapture approach to that conducted by Kamler et al. (2007a, 2007b) to assess what affects this new analysis had on our conclusions. We first built a fully parameterized model that included all 4 variables. Next, we individually removed season (i.e., species*year*location) and year (i.e., species*season*location) to simplify our analysis and focus on our 2 main hypotheses, effects of coyote density and landscape type. Body mass, body length, and relative body mass (body length/body mass) were compared across sexes within landscape types and across landscape types within sexes using Student’s t-tests (Sokal and Rohlf 1999). Statistical analyses were conducted using SAS 9.1 (SAS Institute Inc., Cary, North Carolina).
RESULTS We collected a total 490 scats of swift foxes and 482 scats of coyotes from January 1999 through December 2000. Twenty-six taxonomic groups of food remains were identified: insects, lagomorphs (including Lepus californicus and Sylvil-
69
TABLE 1.—Models used to estimate diets of swift foxes and coyotes in northwestern Texas from January 1999 through December 2000. Modela
QAICc
DQAICc
sp*se*loc sp*loc sp*yr se*yr*loc sp*se yr*loc se+(sp*loc) (sp*se)+loc sp+(se*loc) sp*yr*se yr*se*loc intercept sp*yr*se*loc
1,842.351 1,861.010 1,874.413 1,896.019 1,903.625 1,929.187 1,966.991 1,969.581 1,969.853 1,977.410 2,054.359 2,158.963 2,167.499
0.0000 18.6591 32.0618 53.6685 61.2745 86.8361 124.6402 127.2296 127.5025 135.0588 212.0084 316.6125 325.1485
, , , , , , , , , , , ,
MWb
Kc
0.99991 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
129 33 33 65 65 33 120 122 122 129 129 2 257
a Model notations: sp 5 species, yr 5 year, se 5 season, and loc 5 location. Interactive models are designated by an asterisk (*) between factors, and additive models are designated by a plus symbol (+) between factors. For all models the probability of capture (pi) equals the probability of recapture (ci). b MW 5 model weights estimated using program MARK. c K 5 number of parameters in model.
agus audubonii), rodents, cattle, pronghorns (Antilocapra americana), birds, crop seeds (including corn, wheat, sorghum, and sunflower seed), and vegetation (including grass and forbs). Small rodents identified in scats included Ord’s kangaroo rats (Dipodomys ordii), pocket mice (Perognathus spp. and Chaetodipus spp.), deer mice (Peromyscus spp.), harvest mice (Reithrodontomys spp.), and northern grasshopper mice (Onychomys leucogaster). Large rodents identified in scats included black-tailed prairie dogs (Cynomys ludovicianus), woodrats (Neotoma spp.), hispid cotton rats (Sigmodon hispidus), pocket gophers (Cratogeomys spp.), and ground squirrels (Spermophilus spp.; Appendices I and II). We fit 13 models with 4 unique variables (species, year, season, and location) to the data (Table 1). The only competitive model (i.e., 4 QAICc) contained an interaction between species, season, and location, suggesting that landscape type was the most important predictor of diets of swift foxes and coyotes. This model accounted for more than 99% of the model weights. Our 2nd-best model (DQAICc 5 18.6591) included a species by location interaction. Overall, we interpret these results to indicate that diets varied substantially with season within species and that landscape type was the most important predictor of diets in this study. The best model containing a species by year by location interaction performed poorly (DQAICc 5 53.6685), providing little support for the hypothesis that coyote abundance influenced diets of swift foxes in the study area. Model-averaged parameter estimates indicated that insects and small rodents comprised the majority of the diets of swift foxes (Figs. 1 and 2), whereas the diets of coyotes were more diverse and included insects, small rodents, large rodents, and lagomorphs in approximately equal proportions (Figs. 3 and 4). Our model-averaged estimates also revealed that most variation in diets of both canids was explained by season and location, and annual differences accounted for a relatively
70
JOURNAL OF MAMMALOGY
Vol. 91, No. 1
FIG. 1.—Proportion of insects and small rodents in the diet of swift foxes in fragmented and grassland landscapes in northwestern Texas from winter 1999 to fall 2000. Error bars in figure represent standard errors. See Appendix I for sample sizes.
small amount of variation in the diet. Overall, insects occurred most often in the diet of swift foxes throughout the year, and small mammals commonly occurred in summer and fall (Figs. 1 and 2). Coyote diets were composed primarily of mammals, including small and large rodents and lagomorphs in approximate equal proportions, whereas insects were common during the summer season only (Figs. 3 and 4). Both swift foxes and coyotes included proportionally more insects in their diets on grassland landscapes, consuming more small rodents, large rodents, and lagomorphs on fragmented landscapes (Table 2).
We observed strong seasonality in diets of swift foxes in grasslands. Use of small mammals peaked in the spring of both years, and use of insects peaked in fall (Fig. 1). Seasonality in diets of swift foxes was less predictable in the fragmented landscape, although use of small rodents peaked during summer of both years (Fig. 1). Seasonality in the use of large rodents by swift foxes was less apparent in both landscapes (Fig. 1). For coyotes in grassland, use of insects peaked sharply during summer of both years (Figs. 3 and 4), and this pattern was similar but less pronounced for both small and large rodents (Figs. 3 and 4). In contrast,
February 2010
LEMONS ET AL.—VARIATION IN CANID DIETS
71
FIG. 2.—Proportion of large rodents and lagomorphs in the diet of swift foxes in fragmented and grassland landscapes in northwestern Texas from winter 1999 to fall 2000. Error bars in figure represent standard errors. See Appendix I for sample sizes.
coyotes in the fragmented landscape consumed small rodents most and insects least in winter of both years (Fig. 3). Lagomorphs peaked in the diets of coyotes in fragmented landscapes during spring of both years (Fig. 4). Differences in body size were apparent when testing between landscape types by sex (Fig. 5). Female swift foxes occupying the fragmented landscape had a larger mass (t15 5 4.317, P , 0.001) and a larger relative body mass (t15 5 3.757, P 5 0.002) than those occupying continuous grassland landscapes. No differences were recorded in body length (t15 5 1.214, P 5 0.243) of females between the 2 landscape
types. No differences were recorded for body mass (t15 5 0.140, P 5 0.891), relative body mass (t15 5 20.270, P 5 0.791), or body length (t15 5 0.653, P 5 0.524) for male swift foxes between landscape type.
DISCUSSION Over at least the last half century the most common technique for assessing diets of vertebrates has been collection and analysis of fecal, stomach, or pellet samples, or a combination of these. However, these samples typically
72
JOURNAL OF MAMMALOGY
Vol. 91, No. 1
FIG. 3.—Proportion of insects and small rodents in the diet of coyotes (Canis latrans) in fragmented and grassland landscapes in northwestern Texas from winter 1999 to fall 2000. Error bars in figure represent standard errors. See Appendix II for sample sizes.
contain more than 1 food item per sampling unit (Bonesi et al. 2004; Miyasaka et al. 2003; Pechacek and Kristin 2004; Phillips et al. 2007; Post 2003; Vander Zanden and Vadeboncoeur 2002). Traditional data analyses consist of the Pearson chi-square test for independence, ANOVA, or other similar techniques. These techniques are generally inappropriate because of the multiple responses and correlation among food items (Hurlbert 1984). For example, in a related problem involving ‘‘choose all that apply’’ survey questions, numerous studies (Agresti and Liu 1999, 2001; Bilder et al. 2000; Decady and Thomas 2000; Loughin and Sherer 1998)
have demonstrated that the traditional Pearson chi-square test is inappropriate and can result in tests of hypotheses that do not achieve the stated type I error level. A previously published analysis of data on the diets of swift foxes and coyotes in northwestern Texas using traditional approaches suggested that both coyote density and landscape type were important predictors of diets of swift foxes (Kamler et al. 2007a, 2007b). Our reassessment of these data using the capture–mark–recapture multinomial modeling approach reinforced the idea that landscape type influenced prey use by swift foxes but failed to detect an important effect of coyote
February 2010
LEMONS ET AL.—VARIATION IN CANID DIETS
73
FIG. 4.—Proportion of large rodents and lagomorphs in the diet of coyotes (Canis latrans) in fragmented and grassland landscapes in northwestern Texas from winter 1999 to fall 2000. Error bars in the figure represent standard errors. See Appendix II for sample sizes. TABLE 2.—Proportion of swift fox and coyote scats in which each of 4 prey types occurred in northwestern Texas from winter 1999 to fall 2000. Insects
Small rodentsa
Large rodentsb
Lagomorphs
Grassland landscapes Fragmented landscapes
0.4729 0.3411
0.2906 0.3824
0.0997 0.1082
0.0315 0.0718
Coyotes Grassland landscapes Fragmented landscapes
0.2352 0.1610
0.1604 0.2181
0.1835 0.2140
0.0735 0.1490
Swift foxes
a b
Average body mass , 100 g. Average body mass . 100 g.
density on diets of swift foxes. The results from our capture– mark–recapture models did not support the suggestion from Kamler et al. (2007a) of dietary competition between swift foxes and coyotes, which was likely an artifact of their inappropriate analytical methods. Our analyses also revealed that female but not male swift foxes occupying fragmented landscapes were in better overall condition based on body mass and relative body mass than were swift foxes in areas of continuous grasslands. Female swift foxes are philopatric, whereas males disperse, and the similarity in body size for male foxes captured in the 2 landscapes may have been related to their exposure to a diversity of prey in varied habitats during the dispersal period.
74
Vol. 91, No. 1
JOURNAL OF MAMMALOGY
study completed in northwestern Texas. Our results suggest that the evidence of competition between swift foxes and coyotes reported by Kamler et al. (2007a) may have been an artifact of inappropriate analytical methods. We recommend that quantitative procedures used to analyze diet from scats, pellets, and stomach samples account for the inherent pseudoreplication associated with these types of data. When past diet studies have the potential to alter fundamentally our understanding of biological systems, or have important relevance for management, we recommend that they be revisited using more appropriate analytical methods.
ACKNOWLEDGMENTS S. Jenkins, A. Agresti, and C. Builder provided help and comments throughout this study. Texas Parks and Wildlife, Houston Zoological Society, BWXT Pantex, and the Northern Prairie Research Center provided funding and support. We appreciate the help of the landowners in Sherman and Dallam counties, Texas, and particularly F. Pronger for providing access to his land during the term of this study. We also thank W. Ballard and J. Kamler for assistance in the field and comments on study design and collection of samples.
LITERATURE CITED
FIG. 5.—Mean (6 SD) body length, body mass, and relative body mass (body length/body mass) of male and female swift foxes in continuous grassland and fragmented landscapes in northwestern Texas from winter 1999 to fall 2000. Bars with different letters were significantly different at P , 0.05 using Student’s t-test.
A number of studies have reported that variation in the abundance and availability of prey (particularly small mammals and lagomorphs) strongly influences reproductive success and population size among several species of foxes (Cypher et al. 2000; Harris 1979; Kolb and Hewson 1980; White et al. 1996). Our results indicated that swift foxes consumed a greater diversity of mammalian prey in fragmented compared to continuous grassland landscapes, which contributed to overall better body condition for female foxes. Body condition links to survival and reproduction in foxes (Cypher et al. 2000), and our results therefore suggest the potential for landscape characteristics to have an important influence on the population dynamics of swift foxes. In summary, we suggest that many of the traditional approaches used to investigate diets in wildlife studies are inappropriate by failing to account for the lack of independence when the sample units (scats, pellet, and stomach contents) contain multiple food items. We used multinomial models developed for capture–mark–recapture data to reassess information on the diets of swift foxes and coyotes from a
AGRESTI, A., AND I. LIU. 1999. Modeling a categorical variable allowing for arbitrarily many category choices. Biometrics 55:936– 943. AGRESTI, A., AND I. LIU. 2001. Strategies for modeling a categorical variable allowing multiple category choices. Sociological Methods and Research 29:403–434. ANDERSON, D. R., K. P. BURNHAM, AND G. C. WHITE. 1994. AIC model selection in overdispersed capture–recapture data. Ecology 75:1780–1793. AZEVEDO, F. C. C., V. LESTER, W. GORSUCH, S. LARIVIERE, A. J. WIRSING, AND D. L. MURRAY. 2006. Dietary breadth and overlap among five sympatric prairie carnivores. Journal of Zoology (London) 269:127–135. BADZINSKI, S. S., AND S. A. PETRIE. 2006. Diets of lesser and greater (London) scaup during autumn and spring on the lower Great Lakes. Wildlife Society Bulletin 34:664–674. BILDER, C. R., T. M. LOUGHIN, AND D. NETTLETON. 2000. Multiple marginal independence testing for pick any/c variables. Communications in Statistics: Simulation and Computation 29:1285–1316. BONESI, L., P. CHANIN, AND D. W. MACDONALD. 2004. Competition between Eurasian otter Lutra lutra and American mink Mustela vison probed by niche shift. Oikos 106:19–26. BURNHAM, K. P., AND D. R. ANDERSON. 2002. Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed. Springer-Verlag, New York. CARBYN, L. N., H. J. ARMBRUSTER, AND C. MAMO. 1994. The swift fox reintroduction program in Canada from 1983 to 1992. Pp. 247–271 in Restoration of endangered species: conceptual issues, planning and implementation (M. L. Bowles and C. J. Whelan, eds.). Cambridge University Press, Cambridge, United Kingdom. COVELL, D. F. 1992. Ecology of the swift fox (Vulpes velox) in southeastern Colorado. M.S. thesis, University of Wisconsin, Madison. CYPHER, B. L., ET AL. 2000. Population dynamics of San Joaquin kit foxes at the Naval Petroleum Reserves in California. Wildlife Monographs 145:1–43.
February 2010
LEMONS ET AL.—VARIATION IN CANID DIETS
DECADY, Y. J., AND D. H. THOMAS. 2000. A simple test of association for contingency tables with multiple column responses. Biometrics 56:893–896. EGOSCUE H. J. 1979. Vulpes velox. Mammalian Species. 122:1–5. FLOYD, T. J., J. D. MECH, AND P. A. JORDAN. 1978. Relating wolf scat content to prey consumed. Journal of Wildlife Management 42:528–532. GIER, H. T. 1968. Coyotes in Kansas. Kansas Agricultural Experiment Station Bulletin 393:1–118. GILBERT, B. M. 1990. Mammalian osteology. Missouri Archaeological Society, Columbia. GLEN, A. S., AND C. R. DICKMAN. 2006. Diet of spotted-tailed quoll (Dasyurus maculates) in eastern Australia: effects of season, sex and size. Journal of Zoology (London) 269:241–248. HARRIS, S. 1979. Age-related fertility and productivity in red foxes, Vulpes vulpes, in suburban London. Journal of Zoology (London) 187:195–199. HENKE, S. E. 1992. Effect of coyote removal on the faunal community ecology of a short-grass prairie. Ph.D. dissertation, Texas Tech University, Lubbock. HENKE, S. E., AND F. C. BRYANT. 1999. Effects of coyote removal on the faunal community in western Texas. Journal of Wildlife Management 63:1066–1081. HINES, T. D. 1980. An ecological study of Vulpes velox in Nebraska. M.S. thesis, University of Nebraska, Lincoln. HUGGINS, R. M. 1989. On the statistical analysis of capture experiments. Biometrika 76:133–140. HURLBERT, S. H. 1984. Pseudoreplication and the design of ecological experiments. Ecological Monographs 54:187–211. KAMLER, J. F., W. B. BALLARD, R. L. GILLILAND, P. R. LEMONS, AND K. MOTE. 2003. Impacts of coyotes on swift foxes in northwestern Texas. Journal of Wildlife Management 67:317–323. KAMLER, J. F., W. B. BALLARD, M. C. WALLACE, R. L. GILLILAND, AND P. S. GIPSON. 2007a. Dietary overlap of swift foxes and coyotes in northwest Texas. American Midland Naturalist 158:139–146. KAMLER, J. F., W. B. BALLARD, M. C. WALLACE, AND P. S. GIPSON. 2007b. Diets of swift foxes (Vulpes velox) in continuous and fragmented prairie in northwestern Texas. Southwestern Naturalist 52:504–510. KELLY, B. T. 1991. Carnivore scat analysis: an evaluation of existing techniques and the development of predictive models of prey consumed. M.S. thesis, University of Idaho, Moscow. KELLY, B. T., AND E. O. GARTON. 1997. Effects of prey size, meal size, meal composition, and daily frequency of feeding on the recovery of rodent remains from carnivore scats. Canadian Journal of Zoology 75:1811–1817. KILGORE, D. L. 1969. An ecological study of the swift fox (Vulpes velox) in the Oklahoma Panhandle. American Midland Naturalist 81:512–533. KITCHEN, A. M., E. M. GESE, AND E. R. SCHAUSTER. 1999. Resource partitioning between coyotes and swift foxes: space, time, and diet. Canadian Journal of Zoology 77:1645–1656. KNOWLTON, F. F. 1964. Aspects of coyote predation in south Texas with special reference to white-tailed deer. Ph.D. dissertation, Purdue University, West Lafayette, Indiana. KOLB, H. H., AND R. HEWSON. 1980. A study of fox populations in Scotland from 1971 to 1976. Journal of Applied Ecology 17:7– 19. LINHART, S. B., AND W. B. ROBINSON. 1972. Some relative carnivore densities in areas under sustained coyote control. Journal of Mammalogy 53:880–884.
75
LITVAITIS, J., A. K. TITUS, AND E. M. ANDERSON. 1996. Measuring vertebrate use of terrestrial habitats and foods. Pp. 254–274 in Research and management techniques for wildlife and research (T. A. Bookhout, ed.). 5th ed. Wildlife Society, Bethesda, Maryland. LOCKIE, J. D. 1959. The estimation of food of foxes. Journal of Wildlife Management 23:224–229. LOUGHIN, T. M., AND P. N. SHERER. 1998. Testing for association in contingency tables with multiple column responses. Biometrics 54:630–637. MACKENZIE, D. I., J. D. NICHOLS, G. B. LACHMAN, S. DROEGE, J. A. ROYAL, AND C. A. LANGTIMM. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83:2248–2255. MACKENZIE, D. I., J. D. NICHOLS, J. A. ROYLE, K. H. POLLOCK, L. L. BAILEY, AND J. E. HINES. 2006. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Elsevier, San Diego, California. MATLACK, R. S., P. S. GIPSON, AND D. W. KAUFMAN. 2000. The swift fox in rangeland and cropland in western Kansas: relative abundance, mortality, and body size. Southwestern Naturalist 45:221–225. MIYASAKA, H., M. GENKAI-KATO, N. KUHARA, AND S. NAKANO. 2003. Predatory fish impact on competition between stream insect grazers: a consideration of behaviorally- and density-mediated effects on an apparent coexistence pattern. Oikos 101:511–520. MOORE, T. D., L. E. SPENCER, AND C. E. DUGNOLLE. 1997. Identification of the dorsal guard hairs of some mammals of Wyoming. Wyoming Game and Fish Department Bulletin 14:1–177. MURIE, O. J. 1974. A field guide to animal tracks. Houghton Mifflin, Boston, Massachusetts. OTIS, D. L., K. P. BURNHAM, G. C. WHITE, AND D. R. ANDERSON. 1978. Statistical inference from capture data on closed animal populations. Wildlife Monographs 62:1–135. PECHACEK, P., AND A. KRISTIN. 2004. Comparative diets of adult and young three-toed woodpeckers in a European alpine forest community. Journal of Wildlife Management 68:683–693. PHILLIPS, R. B., C. S. WINCHELL, AND R. H. SCHMIDT. 2007. Dietary overlap of an alien and native carnivore on San Clemente Island, California. Journal of Mammalogy 88:173–180. POST, D. M. 2003. Individual variation in the timing of ontogenetic niche shifts in largemouth bass. Ecology 84:1298–1310. SAMUEL, D. E., AND B. B. NELSON. 1992. Foxes (Vulpes vulpes and allies). Pp. 485–490 in Wild mammals of North America (J. A. Chapman and G. A. Feldhamer, eds.). Johns Hopkins University Press, Baltimore, Maryland. SCHAUSTER, E. R., E. M. GESE, AND A. M. KITCHEN. 2002. Population ecology of swift foxes (Vulpes velox) in southeastern Colorado. Canadian Journal of Zoology 80:307–319. SCOTT-BROWN, J. M., S. HERRERO, AND J. REYNOLDS. 1987. Swift fox. Pp. 432–441 in Wild furbearer management and conservation in North America (M. Nowak, J. A. Baker, M. E. Obbard, and B. Malloch, eds.). Ontario Ministry of Natural Resources, Toronto, Ontario, Canada. SOKAL, R. R., AND F. J. ROHLF. 1999. Biometry. 3rd ed. W. H. Freeman, New York. SOVADA, M. A., C. C. ROY, J. B. BRIGHT, AND J. R. GILLIS. 1998. Causes and rates of mortality of swift foxes in western Kansas. Journal of Wildlife Management 62:1300–1306. SOVADA, M. A., C. C. ROY, AND D. J. TELESCO. 2001. Seasonal food habits of swift fox (Vulpes velox) in cropland and rangeland landscapes in western Kansas. American Midland Naturalist 145:101–111.
76
JOURNAL OF MAMMALOGY
SPAULDING, R., P. R. KRAUSMANN, AND W. B. BALLARD. 2000. Observer bias and analysis of gray wolf diets from scats. Wildlife Society Bulletin 28:947–950. THOMAS, D. L., AND E. J. TAYLOR. 1990. Study designs and tests for comparing resource use and availability. Journal of Wildlife Management 54:322–330. VANDER ZANDEN, M. J., AND Y. VADEBONCOEUR. 2002. Fishes as integrators of benthic and pelagic food webs in lakes. Ecology 83:2152–2161. WEAVER, J. L., AND S. W. HOFFMAN. 1979. Differential detectability of rodents in coyote scats. Journal of Wildlife Management 43:783– 786. WHITE, G. C., AND K. P. BURNHAM. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46:S120–S138.
Vol. 91, No. 1
WHITE, P. J., C. A. V. WHITE, AND K. RALLS. 1996. Functional and numerical responses of kit foxes to a short-term decline in mammalian prey. Journal of Mammalogy 77:370–376. WILLIAMS, B. K., J. D. NICHOLS, AND M. J. CONROY. 2002. Analysis and management of animal populations. Academic Press, San Diego, California. ZAR, J. H. 1999. Biostatistical analysis. 4th ed. Prentice Hall, Upper Saddle River, New Jersey. ZEPPELIN, T. K., AND R. R. REAM. 2006. Foraging habits based on diet of female northern fur seals (Callorhinus ursinus) on the Pribolof Islands, Alaska. Journal of Zoology (London) 270:565–576.
Submitted 14 September 2007. Accepted 20 May 2009. Associate Editor was Rick A. Sweitzer.
APPENDIX I
Mammals Sciuridae Spermophilus spp. Cynomys ludovicianus Geomyidae Heteromyidae Dipodomys spp. Chaetodipus spp. Perognathus spp. Muridae Peromyscus spp. Reithrodontomys spp. Neotoma spp. Sigmodon hispidus Microtus spp. Unknown Muridae Leporidae Bovidae Antilocapridae Unknown mammal Birds Arthropods Crops Sorghum Sunflower seed Vegetation
76.9 11.5 7.7 3.8 11.5 26.9 11.5 0.0 15.4 7.7 0.0 3.8 0.0 0.0 0.0 3.8 11.5 3.8 0.0 0.0 30.8 53.8 0.0 0.0 0.0 11.5
89.5 5.3 0.0 5.3 0.0 10.5 0.0 0.0 10.5 36.8 5.3 10.5 0.0 5.3 0.0 15.8 21.1 0.0 0.0 0.0 21.1 52.6 0.0 0.0 0.0 10.5
100.0 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
83.3 0.0 0.0 0.0 16.7 50.0 0.0 0.0 50.0 16.7 0.0 0.0 0.0 0.0 16.7 0.0 0.0 0.0 0.0 0.0 33.3 83.3 0.0 0.0 0.0 0.0
67.9 7.1 7.1 0.0 10.7 35.7 0.0 3.6 32.1 3.6 0.0 3.6 0.0 0.0 0.0 0.0 3.6 7.1 0.0 0.0 3.6 67.9 0.0 0.0 0.0 7.1
107.7 7.7 0.0 7.7 11.5 50.0 0.0 0.0 50.0 19.2 0.0 0.0 0.0 3.8 7.7 7.7 3.8 0.0 0.0 7.7 7.7 57.7 0.0 0.0 0.0 3.8
Frag. (n 5 26)
Summer 1999
Frag. Cont. (n 5 6) (n 5 28)
Spring 1999
Cont. Frag. Cont. (n 5 26) (n 5 19) (n 5 1)
Winter 1999
Winter 2000
Spring 2000
43.8 3.8 3.8 0.0 7.5 20.0 1.3 3.8 15.0 3.8 0.0 2.5 1.3 0.0 0.0 0.0 3.8 3.8 1.3 0.0 2.5 82.5 2.5 2.5 0.0 15.0
95.0 5.0 0.0 5.0 10.0 30.0 5.0 0.0 25.0 35.0 5.0 10.0 0.0 5.0 5.0 10.0 5.0 0.0 0.0 0.0 35.0 40.0 10.0 10.0 0.0 0.0
70.8 3.1 1.5 1.5 7.7 32.3 4.6 0.0 27.7 13.8 1.5 4.6 1.5 1.5 1.5 3.1 3.1 7.7 0.0 0.0 4.6 70.8 0.0 0.0 0.0 4.6
106.7 3.3 3.3 0.0 6.7 13.3 0.0 0.0 13.3 43.3 3.3 16.7 3.3 0.0 6.7 13.3 13.3 10.0 0.0 3.3 23.3 30.0 10.0 0.0 10.0 10.0
76.1 6.5 6.5 0.0 4.3 54.3 6.5 2.2 45.7 4.3 0.0 2.2 2.2 0.0 0.0 0.0 0.0 4.3 0.0 2.2 6.5 43.5 0.0 0.0 0.0 10.9
82.5 7.5 2.5 5.0 2.5 32.5 2.5 5.0 25.0 20.0 5.0 2.5 2.5 2.5 5.0 2.5 15.0 0.0 0.0 2.5 7.5 30.0 2.5 2.5 0.0 0.0
55.6 0.0 0.0 0.0 3.7 51.9 14.8 11.1 25.9 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 81.5 3.7 3.7 0.0 3.7
105.9 0.0 0.0 0.0 11.8 52.9 17.6 0.0 35.3 17.6 0.0 5.9 0.0 0.0 0.0 11.8 11.8 0.0 0.0 0.0 0.0 58.8 0.0 0.0 0.0 0.0
Frag. (n 5 17)
Summer 2000
Cont. Frag. Cont. Frag. Cont. Frag. Cont. (n 5 80) (n 5 20) (n 5 65) (n 5 30) (n 5 46) (n 5 40) (n 5 27)
Fall 1999
73.5 2.9 2.9 0.0 5.9 11.8 8.8 0.0 2.9 23.5 0.0 2.9 0.0 0.0 0.0 20.6 5.9 2.9 0.0 0.0 5.9 67.6 0.0 0.0 0.0 8.8
95.8 0.0 0.0 0.0 8.3 29.2 4.2 12.5 12.5 25.0 4.2 0.0 0.0 0.0 0.0 20.8 12.5 0.0 0.0 0.0 0.0 58.3 0.0 0.0 0.0 0.0
Cont. Frag. (n 5 34) (n 5 24)
Fall 2000
Frequency of occurrence of prey items in swift fox scats on continuous rangeland (Cont.) and fragmented agriculture–rangeland (Frag.) in northwestern Texas from January 1999 to December 2000. n 5 number of scats.
February 2010 LEMONS ET AL.—VARIATION IN CANID DIETS 77
APPENDIX II
Mammals Sciuridae Spermophilus spp. Cynomys ludovicianus Geomyidae Heteromyidae Dipodomys spp. Chaetodipus spp. Perognathus spp. Muridae Peromyscus spp. Reithrodontomys spp. Neotoma spp. Sigmodon hispidus Microtus spp. Unknown Muridae Leporidae Bovidae Antilocapridae Unknown mammal Birds Arthropods Crops Sorghum Sunflower seed Vegetation
Spring 1999
Summer 1999
Fall 1999
Winter 2000
Spring 2000
Summer 2000
Fall 2000
80.0 0.0 0.0 0.0 20.0 10.0 6.7 0.0 3.3 6.7 0.0 6.7 0.0 0.0 0.0 0.0 10.0 30.0 3.3 0.0 0.0 13.3 36.7 3.3 0.0 33.3
160.7 10.7 3.6 7.1 3.6 7.1 3.6 0.0 3.6 103.6 14.3 14.3 14.3 17.9 32.1 10.7 21.4 3.6 0.0 0.0 21.4 7.1 21.4 3.6 0.0 17.9
89.5 21.1 10.5 10.5 10.5 5.3 0.0 0.0 5.3 5.3 0.0 5.3 0.0 0.0 0.0 0.0 26.3 21.1 0.0 0.0 0.0 26.3 15.8 0.0 0.0 15.8
125.0 18.8 0.0 18.8 6.3 6.3 6.3 0.0 0.0 50.0 0.0 0.0 0.0 0.0 37.5 12.5 31.3 0.0 0.0 0.0 0.0 25.0 18.8 0.0 0.0 18.8
84.4 6.7 4.4 2.2 20.0 24.5 6.7 2.2 15.6 8.9 2.2 0.0 0.0 4.4 2.2 0.0 6.7 17.8 0.0 0.0 6.7 55.6 71.1 33.3 0.0 37.8
73.1 11.5 3.9 7.7 0.0 15.4 11.5 0.0 3.9 7.7 0.0 0.0 3.9 0.0 0.0 3.9 15.4 19.2 0.0 0.0 0.0 34.6 26.9 15.4 0.0 11.5
88.4 2.9 1.5 1.5 15.9 13.1 2.9 1.5 8.7 5.8 0.0 1.5 2.9 0.0 1.5 0.0 18.8 31.9 0.0 0.0 1.5 20.3 44.9 2.9 1.5 40.6
93.6 9.7 3.2 6.5 0.0 9.7 0.0 3.2 6.5 45.2 0.0 12.9 0.0 12.9 19.4 0.0 9.7 16.1 3.2 0.0 6.5 25.8 32.3 6.5 6.5 19.4
126.9 9.8 2.4 7.3 22.0 26.8 2.4 0.0 24.4 17.1 0.0 2.4 2.4 0.0 4.9 7.3 9.8 34.2 0.0 0.0 2.4 14.6 19.5 0.0 0.0 19.5
85.7 0.0 0.0 0.0 9.5 0.0 0.0 0.0 0.0 23.8 0.0 0.0 0.0 4.8 19.1 0.0 33.3 14.3 4.8 0.0 9.5 4.8 28.6 4.8 9.5 14.3
118.8 18.8 6.3 12.5 18.8 31.3 12.5 0.0 18.8 12.5 0.0 6.3 0.0 0.0 6.3 0.0 6.3 25.0 6.3 0.0 6.3 6.3 6.3 0.0 0.0 6.3
72.7 9.1 9.1 0.0 4.6 13.6 0.0 0.0 13.6 13.6 4.6 0.0 0.0 9.1 0.0 0.0 22.7 9.1 0.0 0.0 0.0 45.5 22.7 0.0 0.0 22.7
56.5 0.0 0.0 0.0 26.1 17.4 8.7 4.4 4.4 4.4 0.0 0.0 0.0 0.0 0.0 4.4 0.0 4.4 0.0 0.0 13.0 78.3 17.4 0.0 0.0 17.4
72.7 0.0 0.0 0.0 13.6 0.0 0.0 0.0 0.0 40.9 0.0 4.6 13.6 13.6 9.1 0.0 18.2 0.0 0.0 0.0 0.0 18.2 27.3 0.0 0.0 27.3
81.1 0.0 0.0 0.0 13.5 21.6 5.4 10.8 5.4 13.5 0.0 0.0 2.7 0.0 0.0 10.8 10.8 8.1 0.0 2.7 2.7 56.8 10.8 2.7 2.7 5.4
105.0 30.0 5.0 25.0 10.0 0.0 0.0 0.0 0.0 40.0 5.0 5.0 0.0 15.0 10.0 5.0 20.0 0.0 0.0 0.0 5.0 15.0 20.0 5.0 0.0 15.0
Cont. Frag. Cont. Frag. Cont. Frag. Cont. Frag. Cont. Frag. Cont. Frag. Cont. Frag. Cont. Frag. (n 5 30) (n 5 28) (n 5 19) (n 5 16) (n 5 45) (n 5 26) (n 5 69) (n 5 31) (n 5 41) (n 5 21) (n 5 16) (n 5 22) (n 5 23) (n 5 22) (n 5 37) (n 5 20)
Winter 1999
Frequency of occurrence of prey items in coyote scats on continuous rangeland (Cont.) and fragmented agriculture–rangeland (Frag.) in northwestern Texas from January 1999 to December 2000. n 5 number of scats.
78 JOURNAL OF MAMMALOGY Vol. 91, No. 1
