models developed from δ13c and δ15n of skin tissue ... · john r. finnerty & thomas h. kunz2,...

12
13 (1): 11-22 (2006) Dietary habits are important for understanding the ecology and behaviour of animals and for assessing their ecological and economic importance in both natural and anthropogenically impacted ecosystems. Spatial and tem- poral variation in local food resources can have profound influences on daily and seasonal movements, reproduction, intra- and interspecific interactions, and population density. Biochemical processes that discriminate against the heavy isotope of carbon in plants or that result in nitro- gen-enriched soils have proven useful in assessing diets of Models developed from δ 13 C and δ 15 N of skin tissue indicate non-specific habitat use by the big brown bat (Eptesicus fuscus) 1 James C. SULLIVAN & Kendra J. BUSCETTA, Department of Biology, Boston University, Boston, Massachusetts 02215, USA. Robert H. MICHENER, Department of Biology and Stable Isotope Laboratory, Boston University, Boston, Massachusetts 02215, USA. John O. WHITAKER, Jr., Department of Ecology and Organismal Biology, Indiana State University, Terre Haute, Indiana 47809, USA. John R. FINNERTY & Thomas H. KUNZ 2 , Department of Biology, Boston University, Boston, Massachusetts 02215, USA, e-mail: [email protected] Abstract: Stable isotopes can be used to evaluate trophic relationships, nutrient state, and temporal and spatial variation in diet, food webs, and behaviour both within and between species. Here we describe the development and application of models to predict habitat use of a common insectivorous bat (Eptesicus fuscus) based upon δ 13 C and δ 15 N signatures of skin tissue. We used a 42-specimen sample collected from three well-characterized ecogeographic regions, disparate both in photosynthetic mechanism and fertilizer use, to generate the models. Significant univariate differences between these three sites in terms of δ 13 C (F 2, 39 = 112.92, P < 0.0001) and δ 15 N (F 2, 39 = 97.06, P < 0.0001), and multivariate significance of both variables (Wilk’s λ = 0.032, F 4, 76 = 87.02, P < 0.0001), made it possible to develop three predictive models using Fisher’s linear discriminant functions: 1) a model predicting if bats forage in C 3 or mixed C 3 /C 4 sites, 2) a model predicting if bats forage in agricultural areas, and 3) a combined model using both variables to predict specific habitat use. We present the results of model application to an independent dataset of 329 bats sampled from 10 states that included a broad range of δ 13 C (-26.53‰ ≤ δ 13 C ≤ -17.20‰) and δ 15 N (6.36‰ ≤ δ 15 N ≤ 15.60‰) signatures. We validated the use of skin tissue samples (from wing membranes) in the model by comparing the sites used for model development across five tissue types, selecting skin samples for model development due to consistently low variance within this tissue type. Our results indicate non- specific habitat-use by big brown bats. Keywords: conservation, foraging habitat, insectivorous bats. Résumé : Les isotopes stables peuvent être utilisés pour examiner les relations trophiques, l'état nutritionnel et les variations temporelles et spatiales dans la diète, les chaînes alimentaires et le comportement à l'intérieur d'une même espèce ou entre les espèces. Nous décrivons ici le développement et l'application de modèles prédictifs de l'utilisation de l'habitat chez une chauve-souris insectivore commune (Eptesicus fuscus) basés sur les signatures δ 13 C et δ 15 N de la peau. Pour générer les modèles, nous avons utilisé un échantillon de 42 spécimens récoltés dans trois régions écogéographiques très distinctes au niveau du mécanisme photosynthétique et de l'utilisation de fertilisants. Des différences univariées significatives entre les trois sites en termes de δ 13 C (F 2, 39 = 112.92, P < 0.0001) et δ 15 N (F 2, 39 = 97.06, P < 0.0001) et la significativité multidimensionnelle des deux variables (Wilk’s λ = 0.032, F 4, 76 = 87.02, P < 0.0001) ont rendu possible le développement de trois modèles prédictifs utilisant les fonctions discriminantes linéaires de Fischer : 1) un modèle prédisant si les chauves-souris se nourrissent dans des sites exclusivement C 3 ou mélangés C 3 /C 4 ; 2) un modèle prédisant si les chauve-souris se nourrissent dans des secteurs agricoles; et 3) un modèle synthétique utilisant les deux variables pour prédire l'utilisation d'habitats spécifiques. Nous présentons les résultats de l'application des modèles sur un échantillon indépendant de 329 chauves-souris provenant de 10 états et présentant de grandes variations dans les signature δ 13 C (-6.53 ‰ ≤ δ 13 C ≤ -17.20 ‰) et δ 15 N (6.36 ‰ ≤ δ 15 N ≤ 15.60 ‰). Nous avons validé l'utilisation d'échantillons de peau (provenant des membranes des ailes) dans les modèles en comparant, pour cinq types de tissu, les sites utilisés pour le développement des modèles. La peau a été le tissu sélectionné pour développer les modèles à cause de sa faible variance. Nos résultats démontrent que la sérotine des maisons n'utilise pas des habitats spécifiques. Mots-clés : chauves-souris insectivores, conservation, habitat d 'alimentation. Nomenclature: Koopman, 1993. Introduction 1 Rec. 2005-03-29; acc. 2005-05-24. Associate Editor: Don Thomas. 2 Author for correspondence.

Upload: others

Post on 16-Aug-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

13 (1): 11-22 (2006)

Dietary habits are important for understanding the ecology and behaviour of animals and for assessing their ecological and economic importance in both natural and

anthropogenically impacted ecosystems. Spatial and tem-poral variation in local food resources can have profound influences on daily and seasonal movements, reproduction, intra- and interspecific interactions, and population density.

Biochemical processes that discriminate against the heavy isotope of carbon in plants or that result in nitro-gen-enriched soils have proven useful in assessing diets of

Models developed from δ13C and δ15N of skin tissue indicate non-specific habitat use by the big brown bat (Eptesicus fuscus)1

James C. SULLIVAN & Kendra J. BUSCETTA, Department of Biology, Boston University, Boston, Massachusetts 02215, USA.

Robert H. MICHENER, Department of Biology and Stable Isotope Laboratory, Boston University, Boston, Massachusetts 02215, USA.

John O. WHITAKER, Jr., Department of Ecology and Organismal Biology, Indiana State University, Terre Haute, Indiana 47809, USA.

John R. FINNERTY & Thomas H. KUNZ2, Department of Biology, Boston University, Boston, Massachusetts 02215, USA, e-mail: [email protected]

Abstract: Stable isotopes can be used to evaluate trophic relationships, nutrient state, and temporal and spatial variation in diet, food webs, and behaviour both within and between species. Here we describe the development and application of models to predict habitat use of a common insectivorous bat (Eptesicus fuscus) based upon δ13C and δ15N signatures of skin tissue. We used a 42-specimen sample collected from three well-characterized ecogeographic regions, disparate both in photosynthetic mechanism and fertilizer use, to generate the models. Significant univariate differences between these three sites in terms of δ13C (F2, 39 = 112.92, P < 0.0001) and δ15N (F2, 39 = 97.06, P < 0.0001), and multivariate significance of both variables (Wilk’s λ = 0.032, F4, 76 = 87.02, P < 0.0001), made it possible to develop three predictive models using Fisher’s linear discriminant functions: 1) a model predicting if bats forage in C3 or mixed C3/C4 sites, 2) a model predicting if bats forage in agricultural areas, and 3) a combined model using both variables to predict specific habitat use. We present the results of model application to an independent dataset of 329 bats sampled from 10 states that included a broad range of δ13C (-26.53‰ ≤ δ13C ≤ -17.20‰) and δ15N (6.36‰ ≤ δ15N ≤ 15.60‰) signatures. We validated the use of skin tissue samples (from wing membranes) in the model by comparing the sites used for model development across five tissue types, selecting skin samples for model development due to consistently low variance within this tissue type. Our results indicate non-specific habitat-use by big brown bats.Keywords: conservation, foraging habitat, insectivorous bats.

Résumé : Les isotopes stables peuvent être utilisés pour examiner les relations trophiques, l'état nutritionnel et les variations temporelles et spatiales dans la diète, les chaînes alimentaires et le comportement à l'intérieur d'une même espèce ou entre les espèces. Nous décrivons ici le développement et l'application de modèles prédictifs de l'utilisation de l'habitat chez une chauve-souris insectivore commune (Eptesicus fuscus) basés sur les signatures δ13C et δ15N de la peau. Pour générer les modèles, nous avons utilisé un échantillon de 42 spécimens récoltés dans trois régions écogéographiques très distinctes au niveau du mécanisme photosynthétique et de l'utilisation de fertilisants. Des différences univariées significatives entre les trois sites en termes de δ13C (F2, 39 = 112.92, P < 0.0001) et δ15N (F2, 39 = 97.06, P < 0.0001) et la significativité multidimensionnelle des deux variables (Wilk’s λ = 0.032, F4, 76 = 87.02, P < 0.0001) ont rendu possible le développement de trois modèles prédictifs utilisant les fonctions discriminantes linéaires de Fischer : 1) un modèle prédisant si les chauves-souris se nourrissent dans des sites exclusivement C3 ou mélangés C3/C4; 2) un modèle prédisant si les chauve-souris se nourrissent dans des secteurs agricoles; et 3) un modèle synthétique utilisant les deux variables pour prédire l'utilisation d'habitats spécifiques. Nous présentons les résultats de l'application des modèles sur un échantillon indépendant de 329 chauves-souris provenant de 10 états et présentant de grandes variations dans les signature δ13C (-6.53 ‰ ≤ δ13C ≤ -17.20 ‰) et δ15N (6.36 ‰ ≤ δ15N ≤ 15.60 ‰). Nous avons validé l'utilisation d'échantillons de peau (provenant des membranes des ailes) dans les modèles en comparant, pour cinq types de tissu, les sites utilisés pour le développement des modèles. La peau a été le tissu sélectionné pour développer les modèles à cause de sa faible variance. Nos résultats démontrent que la sérotine des maisons n'utilise pas des habitats spécifiques.Mots-clés : chauves-souris insectivores, conservation, habitat d 'alimentation.

Nomenclature: Koopman, 1993.

Introduction

1Rec. 2005-03-29; acc. 2005-05-24. Associate Editor: Don Thomas.2Author for correspondence.

Sullivan et al.: Stable iSotope derived habitat-uSe modelS

1212

animals because plants are basal to food webs and may be differentiated isotopically (Post, 2002). Animal feces and the natural fertilizers produced from them (δ15N: 10-20‰) are generally enriched in 15N relative to artificial fertilizers (δ15N: ~ 0‰) (Handley & Raven, 1992; Macko & Ostrom, 1994; Townsend, Young & Macko, 2003). Although plant nitrogen signatures are depleted relative to soil, the higher signature imparted by the use of natural fertilizers can be used as a marker (Denton et al., 2001).

Variable efficiency of incorporation of the heavy iso-tope of carbon (13C) during photosynthesis has also proven useful in studies of dietary habits. C3, C4, and CAM path-ways vary in terms of the enzymes involved in each and of the timing and physical locations of carbon fixation relative to the timing and location of gas exchange with the atmo-sphere. Because of differences in these pathways, character-istic 13C:12C ratios are found in plant tissues, with C3 plants being depleted in carbon (δ13C = -34 to -22‰) relative to CAM and C4 plants (δ13C = - 20 to -10‰) (Gannes, del Rio & Koch, 1998; Dawson et al., 2002).

The resulting natural variation in abundance of sta-ble isotopes has increasingly become used to determine diets and foraging behaviour of a wide range of organ-isms (reviewed in Hobson, 1999; Peterson & Fry, 1987; Rubenstein & Hobson, 2004), including bats. The purpose of these bat studies has generally been to investigate the relative importance of insects and fruits in diets of frugivo-rous bats (Fleming, 1995; Herrera et al., 2001a,b; Herrera et al., 2002), to investigate seasonal movement patterns and geographic variations in diet (Fleming, Nuñez & Sternberg, 1993; Herrera, Fleming & Findley, 1993; Herrera et al., 2001a; Cryan et al., 2004), or to study the relative contribu-tion of C3 and C4/CAM-based food webs in insectivorous bats through an analysis of guano deposits (DesMarais et al., 1980; Mizutani, McFarlane & Kabaya, 1992a,b).

The big brown bat, Eptesicus fuscus, is of considerable ecological importance because of its ubiquitous occur-rence and extensive geographic range in Central and North America, where it forages on insects in a variety of habitats, including agricultural lands, forested regions, wetlands, des-erts, and urban areas (Kurta & Baker, 1990). Understanding the dietary habits of this and other insectivorous bats has largely been limited to analysis of stomach and fecal con-tents (Black, 1972; Whitaker, 1993; 1995), with the use of fecal analysis increasing because it does not require sacri-ficing bats (Kunz & Whitaker, 1983; Kunz, 2004). Based on gut content and fecal analyses, big brown bats are thought to feed mostly on coleopterans, hemipterans, and homopter-ans (Phillips, 1966; Whitaker, 1993; 1995), although other insects are also eaten (Black, 1972). Unfortunately, this methodology provides little or no information on habitats used while foraging (Brigham, 1991).

To further evaluate the feeding ecology of big brown bats, we have developed a predictive model using carbon and nitrogen stable isotope signatures of skin tissue. The purpose of the model is to classify bats as foraging in C3 or mixed C3/C4 habitat and nitrogen-enriched or nitrogen-depleted food webs (the terms “enriched” and “depleted” here are used in a relative context). We apply this model to a large sample of bats from regions in the United States

disparate in terms of photosynthetic pathways and applica-tion of natural fertilizers. Because stable isotope signatures are known to vary among tissue types (Tieszen et al., 1983; Roth, 2003; Voigt et al., 2003; Voigt & Matt, 2004), we first validated the use of skin tissue in model development by a comparison of signatures of feces, whole blood, blood plasma, hair, and skin biopsies across the sites used for model development. We attempted to place a finer scale on our models by testing the effect of precipitation on carbon signatures in a subset of samples collected from within a precipitation gradient in central and eastern Kansas.

An ability to develop foraging models of the big brown bat should be an important tool in bat conservation. Moreover, habitat-use models should make it possible to evaluate the importance of this and other insectivorous bat species with respect to insect pest control, in both agricul-tural and developed areas.

MethodsStudy areaS and Sample collection

Data used for model development were collected in June and August of 2003. We compared multiple tissue types (hair, skin, whole blood, feces, and sometimes blood plasma) from each bat sampled at the following three well-characterized sites:

1) a C3 forested site free from anthropogenic fertil-ization, from which samples were collected in June 2003 (Hillsborough County, New Hampshire);

2) a C3 forested site free from anthropogenic fertiliza-tion with C4 cropland ~ 3 km away from the site of sample collection, from which samples were collected in August 2003 (Crane Naval Depot, Martin County, Indiana; Figure 1a); and

3) an agricultural area from which bats were collected in August 2003 (Prairie Creek, Vigo County, Indiana; Figure 1b; Whitaker, 2004). At this site, bats were collected from within a ~ 0.2-km2 urban area set inside of a 41-km2 tract that was composed of ~ 20% forested region and ~ 80% agricultural region, as quantified from aerial photographs using WCIF ImageJ, version 1.34h (Wright Cell Imaging Facility, University Health Network, Toronto, Ontario). The primary agricultural crops at this site were corn (C4) and soybean (C3).

Foraging and/or drinking bats were captured using 2.5- × 12-m mist nets set over a stream (Martin, Indiana) or by hand removal from maternity roosts after bats returned from foraging around midnight (Vigo, Indiana; Hillsborough, New Hampshire). Upon capture, bats were placed in separate cotton bags and maintained overnight to collect accumulated feces. A portion of the feces sample (~2 μg) was used for stable isotope analysis, whereas the remainder was used to compare diets by keying out insect remains as described by Whitaker (2004). The total fecal production from a given bat each night was treated as one sample. Each sample was examined using 10-70x magni-fication with a dissecting scope (Olympus America SZH, Melville, New York, USA). Percentage volume of each prey item was estimated visually and then grouped by order to

develop relative contributions of each insect taxon to the diets of each big brown bat at each site.

Wing biopsies were taken from the chiropatagium using a standard 4 mm dermal biopsy punch (Miltex GmbH, Tuttlingen, Germany; Worthington Wilmer & Barratt, 1996). Subsequently, using a 25-gauge needle, approximately 30 µL of blood was obtained by venipuncture from a caudal vein of each bat and collected into a microcapillary tube (Voigt et al., 2003). Blood was then centrifuged to separate plasma from formed elements. Tissue scissors were used to col-lect hair samples from the mid-dorsal region of each bat. Individuals were released at the site of capture after being processed. Tissue and fecal samples were maintained on ice or gel coolants in the field and subsequently transferred to a -20 °C freezer. Our tissue sampling techniques caused minimal stress to the animals, and based on our previous observations, wing membranes regenerated from biopsy punches within 3 weeks. All protocols used in this study were approved by Boston University’s Animal Care and Use Committee.

For each site, we compared tissue types to determine if significant differences were present and quantified the relationships between tissue types. We also compared each tissue type sampled across the three well-characterized sites to determine if carbon and nitrogen signatures varied con-sistently as a function of expected photosynthetic pathways and fertilizer use.

Data from the 42 specimens at these sites were used to develop three models of foraging behaviour. The models were developed using discriminant analyses (SAS ver-sion 6.12, SAS Institute, Cary, North Carolina, USA) after appropriate relationships were determined to be significant. These models were then used to evaluate isotope signatures from skin biopsies collected from bat specimens provided

by state rabies diagnostic laboratories (n = 329). Whole ani-mal, non-rabid specimens provided by the laboratories were shipped to us on gel coolants or dry ice and stored in our laboratory at -20 °C until samples were dried and prepared for analysis (Hobson, Gibbs & Gloutney, 1997). Samples were collected from more than 102 counties across 10 states (Table I) between April and August. Data on county and date of collection were provided by the state laboratories for most specimens and are available from the corresponding author.

Water uSe efficiency (Wue) in plantS aS a meanS of determining foraging behaviour

To test for an effect of water use efficiency in plants on bat tissue signatures, we used 19 specimens collected along a precipitation gradient (50-110 cm) in central and eastern Kansas and analyzed these samples based upon historical precipitation information (Policy Research Institute, 2004; Spatial Climate Analysis Service, 2004).

Stable iSotope analySiS and StatiSticS

Tissue samples were removed from the freezer and placed in a 60 °C drying oven for 24 to 48 h. Samples were weighed on a Sartorius microbalance (Sartorius AG, Goettingen, Germany) and loaded into tin capsules for analysis. Results are reported using the standard del nota-tion with units of “per mil” (‰) (Craig, 1953):

[1]

with X referring to the heavy isotopic species of either car-bon (13C) or nitrogen (15N) and R referring to the ratio of the heavy (13C or 15N) to the light (12C or 14N, respectively) isotope. Results were analyzed statistically using both SAS (version 6.12, SAS Institute, Cary, North Carolina, USA) and JMP (version 5.0.1, SAS Institute, Cary, North

ÉcoScience, vol. 13 (1), 2006

13

figure 1. Sites used for model development. a) A C3 forested site (Martin, Indiana) abutting C4 cropland (outside frame) and b) a C3/C4 site consisting of both C3 forest (~20%) and C3/C4 crops (~80%, Vigo, Indiana) are pictured. The third site (Hillsborough, New Hampshire) consists exclusively of C3 forest (not pictured).

δX (‰) = (RSample/RStandard –1) * 1000

Sullivan et al.: Stable iSotope derived habitat-uSe modelS

14

Carolina, USA). Statistical comparisons included Student’s t-test and post hoc Tukey’s honest significant difference (Tukey-Kramer HSD) tests with a modification to control for sample size (Kramer, 1956). Wilk’s λ was used to test for multivariate significance with the Bonferroni procedure applied to control type I error. The Shapiro-Wilk statistic (W) was used to test assumptions of normality (Shapiro & Wilk, 1965).

To compare diet at each site, the percent contribution of each identified order of insect was quantified by percent volume and arcsine square-root transformed. If transformed data were distributed normally, the three sites were com-pared for each order of insects using one-way ANOVAs with the Bonferroni procedure applied to control type I error (SAS, version 6.12). If the transformed data were not normally distributed, Chi-Square analysis was performed to compare sites based on the presence and absence of an order in feces in each sample at each site (JMP version 5.0.1).

Sample analySiS

Samples were combusted and analyzed using a Fisons NA1500 elemental analyzer and a Finnigan continuous flow system, coupled to a Delta-S isotope ratio mass spec-trometer at Boston University’s Stable Isotope Laboratory.

Carbon and nitrogen signatures of samples were calibrated to the standards of Vienna-Pee Dee Belemnite (Coplen, 1996) and atmospheric nitrogen (Mariotti, 1984), respec-tively. Instrument precision of replicate samples was less than 0.2‰ for both δ13C and δ15N (Jardine & Cunjak, 2005). Ninety-two specimens were sampled in duplicate as a control to determine tissue consistency. Percent relative standard deviation (%RSD) was calculated by dividing the standard deviation of the duplicates by the mean and multi-plying by 100. For δ13C values, the mean %RSD was 1.17 ± 0.17 SE and was less than 3.00 for 97.5% of duplicate samples. For δ15N values, the mean %RSD was 1.48 ± 0.16 and was less than 5.00 for 97.5% of duplicate samples.

Results

compariSon of diet by fecal analySiS

Bats sampled from Hillsborough, New Hampshire con-sumed almost entirely coleopteran species, whereas bats from both Martin and Vigo, Indiana consumed primarily coleop-teran species and secondarily hemipteran species. Significant differences occurred between sites in percent consumption of coleopteran and hemipteran species and in the distribution of other orders of insects consumed (Table II).

compariSon of SignatureS of Whole blood, blood plaSma, fecal matter, hair, and Skin

A general trend is evident in the three locations used for model development, in which signatures of both car-bon and nitrogen increased in bat tissues as follows: feces < whole blood < hair < skin < blood plasma. For each site and both isotopic signatures, feces was statistically distinct from all other tissue types (Table III) and also had the larg-est variance of sample types that were analyzed. Differences between locations in terms of statistical groupings of other tissue samples were evident (Figure 2).

table i. Means and variation in δ13C and δ15N of skin tissue by state. This table provides sample size and mean δ13C and δ15N signatures with standard errors for each group applied to or included in our model (*: indicates states used for model development). If only one or two counties were sampled for a state, it is listed below the state. Where more than two counties were sampled, the total number of counties is given (most counties are available from the authors). Analyses of variance were conducted on both δ13C and δ15N and found significant differences between states (F12, 359 = 52.18, P < 0.0001; F12, 359 = 14.48, P < 0.0001, respectively). Groupings deter-mined by post hoc Tukey-Kramer HSD tests are superscripted after the means of both δ13C and δ15N.

State δ13C δ15N Counties n Mean SE Mean SEColorado 20 -21.40B 0.25 11.72a 0.32 Larimer*Indiana Martin 14 -20.84B 0.17 8.06e 0.12 Vigo 16 -20.37B 0.38 10.12bcde 0.18Kansas 6 20 -20.99B 0.24 10.51ab 0.16Kentucky Unknown 42 -22.57C 0.18 8.82de 0.14Montana 12 26 -23.79DE 0.22 10.81ab 0.67Nebraska 10 25 -18.83A 0.2 10.53ab 0.13*New Hampshire Hillsborough 12 -24.79E 0.13 8.11de 0.13New Jersey 19 34 -22.88CD 0.17 9.10cde 0.23New Mexico 4 5 -20.58ABB 1.52 9.14bcde 0.37New York 28 50 -23.88EF 0.17 9.33cd 0.19Rhode Island 6 79 -23.70E 0.13 9.32cd 0.1Wisconsin 16 28 -23.77DE 0.24 10.00bc 0.37

table ii. Summary of fecal analysis. "n" indicates the number of sampled individuals that had presence of an insect order in feces. Average % by volume indicates the percentage of the total feces of each insect order and includes all bat feces collected at each site. One-way analyses of variance were used to compare the three sites in terms of % consumption of Order Coleoptera and Order Hemip-tera using data from fractions transformed by an arcsine square-root function. Post hoc Tukey-Kramer HSD tests indicated that Martin and Vigo, IN are significantly different from Hillsborough, NH. The site comparisons for all other insect Orders use Chi-Square analysis. (* P ≤ 0.05, **P ≤ 0.001, *** P ≤ 0.0001, ns = non significative).

Martin, IN Vigo, IN Hillsborough, NH (N = 14) (N = 16) (N = 12) Average % n Average % n Average % by n by volume by volume by volumeAraneaens 0 0 0 0 0.83 ± 1.23 1Coleoptera*** 65.29 ± 6.09 14 68.06 ± 5.69 16 97.08 ± 6.58 12Diptera* 1.85 ± 1.41 5 0 0 0 0Hemiptera*** 25.14 ± 6.11 5 30.94 ± 5.71 16 0.42 ± 6.60 1Hymenopterans 0.86 ± 0.51 2 0.63 ± 0.47 2 0 0Lepidopterans 2.79 ± 0.88 5 0.38 ± 0.82 2 0.42 ± 0.95 1Neuroptera** 2.64 ± 1.96 6 0 0 0 0Trichopterans 1.43 ± 2.48 1 0 0 0 0

ÉcoScience, vol. 13 (1), 2006

15

Correlations were calculated for a comparison of all tissue types grouped across sites, and regressions were fit to model enrichment in tissues from skin tissue samples. A sig-nificant correlation was found for each tissue type for both carbon and nitrogen signatures (Figure 3 and data not pre-sented). Although significant, feces showed low correlation with incorporated tissue (0.38 ≤ r ≤ 0.75), whereas the signa-tures of all other tissue types were highly correlated (0.77 ≤ r ≤ 0.96). A significant and predictive relationship was found between skin tissue signatures and signatures of hair, blood, and plasma for both carbon and nitrogen (Figure 3).

model development (1): δ13c aS a function of ecoregion

We compared sites using δ13C signatures and detected significant differences between sites for each tissue type

tested. Each tissue type showed significant differences between bats sampled from Hillsborough, New Hampshire (NH) and bats sampled from Vigo County, Indiana (IN). Disparities were present in statistical groupings when Vigo and Martin Counties, IN were compared using different tis-sue types (Figure 2).

Of the tissues tested, skin had the smallest variance, and signatures were highly correlated with signatures from all other tissues except feces (Figure 3). Thus, skin samples were used to generate a model to classify bats as belonging to C3 or C3/C4 food webs as a function of δ13C signatures, with samples from Hillsborough, NH entered into the model as a group indicative of a C3 food web and Vigo and Martin, IN samples entered into the model as a group indicative of a C3/C4 food web. The use of these groups was validated

table iii. Tukey-Kramer HSD tests of tissue signatures by site. Groups are results of Tukey-Kramer HSD test groupings. Groups are presen-ted in increasing order of signature for both δ13C and δ15N as are tissue types within groups. Key for tissues: a, fecal matter; b, whole blood; c, hair; d, skin; e, plasma.

δ13C δ15NCounty, state Tissues sampled df F P Groups df F P GroupsVigo, IN a, b, c, d, e 4, 66 27.32 < 0.0001 1) a 4, 63 25.70 < 0.0001 1) a 2) b, d, c, e 2) b, c 3) c, d 4) d, eMartin, IN a, b, c, d 3, 51 119.64 < 0.0001 1) a 3, 51 41.43 < 0.0001 1) a 2) b, c 2) b, c, d 3) c, d Hillsborough, NH a, b, c, d, e 4, 53 22.04 < 0.0001 1) a 4, 53 67.09 < 0.0001 1) a 2) b, d 2) b, e 3) c, e 3) e, c 4) d

figure 2. Mean (± SE) signatures of each tissue type by site. Sample size is presented above δ15N signature bars and below δ13C signature bars. Letters above and below each signature bar indicate results of post hoc Tukey-Kramer HSD tests and are unique for each tissue type and signature. For example, the groupings above δ15N for feces present the results of a one-way ANOVA and HSD test between sites for that tissue type signature.

Sullivan et al.: Stable iSotope derived habitat-uSe modelS

16

by a one-way ANOVA and post hoc Tukey-Kramer HSD test, indicating that samples from Vigo, IN (n = 16, δ13C = -20.37 ± 0.38‰) and Martin, IN (n = 14, δ13C = -20.84 ± 0.17‰) were statistically distinct from Hillsborough, NH (n = 12, δ13C = -24.79 ± 0.13‰), but not distinct from one another (F2, 39 = 112.92, P < 0.0001).

Logistic regression could not be performed because the range of the C3 and C3/C4 groups had no data points that overlapped in terms of δ13C signature (Figure 4), thus pre-

venting an estimation of the maximum likelihood (Albert & Anderson, 1984) . Instead, a linear discriminant function was derived using Fisher's classification functions:

[1.1]

[1.2]

[1.3]

figure 3. Least squares regressions of δ13C (a-d) and δ15N (e-h) of various tissue types against signatures of skin tissues with 95% confidence intervals. While each relationship is significant, the relationship between feces and skin is poorly predictive in the case of both a) carbon and e) nitrogen signatures. All other relationships are significant and highly predictive, as indicated by the Pearson product moment correlations. Legend: triangles, Martin County, Indiana; circles, Vigo County, Indiana; squares, Hillsborough County, New Hampshire.

f0 = -301.35 – 29.24*δ13C

f1 = -437.82 – 35.22*δ13C

L*δ13

C = f1 – f0

ÉcoScience, vol. 13 (1), 2006

17

with the following classification rules: If L*δ13C > 0, then the sample is classified as a C3 forager. If L*δ13C < 0, then the sample is classified as a C3/C4 forager.

Resubstitution was used to test the accuracy of the model by developing the model, excluding a single observation, and testing the accuracy of the model on that datum point. This procedure was repeated iteratively for each observation and indicated 100% accuracy within the model dataset.

model development (2): δ15n aS a function of ecoregion

We compared sites using δ15N signatures and detect-ed significant differences between sites for each tissue type tested. Post hoc Tukey-Kramer HSD tests for each tissue type indicated that samples from Martin, IN and Hillsborough, NH were significantly different from the Vigo, IN samples (Figure 2).

Skin samples had the smallest variance for δ15N signa-tures, and signatures were highly correlated with signatures from all other tissues except feces (Figure 3). Thus, skin samples were used to generate a model to classify bats into “15N-enriched food webs” or “15N-depleted food webs”. Samples from the agricultural site in Vigo, IN were entered into the model as a group indicative of “15N-enriched food webs” and samples from Martin, IN and Hillsborough, NH were entered into the model as a group indicative of “15N-depleted food webs”. The use of these groups was validated by a one-way ANOVA and post hoc Tukey-Kramer HSD test, indicating that skin samples from Martin, IN (n = 14, δ15N = 8.06 ± 0.12‰) and Hillsborough, NH (n = 12, δ15N = 8.11 ± 0.13‰) were significantly different from Vigo, IN samples (n = 16, δ15N = 10.11 ± 0.18‰), but not distinct from one another (F2, 39 = 97.06, P < 0.0001).

Logistic regression could not be performed because the range of the groups had no data points that overlapped in

terms of a δ15N signature (Figure 4). A linear discriminant function, using Fisher's classification functions, was derived:

[2.1]

[2.2]

[2.3]

with the following classification rules: If L*δ15N > 0, then the bat is classified as a forager in 15N-enriched food webs. If L*δ15N < 0, then the bat is classified as a forager in 15N-depleted food webs.

Resubstitution was used to test the accuracy of Model 2 as described for Model 1 and indicated 100% accuracy within the model dataset.

model development (3): multivariate model

The three sites used for model development can be classified as: 1) C3/C4, 15N enriched (Vigo, IN), 2) C3/C4, 15N depleted (Martin, IN), and 3) C3, 15N depleted (Hillsborough, NH). Multivariate significance with δ13C and δ15N each being a function of site (Wilk’s λ = 0.032, F4, 76 = 87.02, P < 0.0001, Figure 4) allowed development of a multivariate discriminant function using Fisher's clas-sification functions:

[3.1]

[3.2]

[3.3]

with the following classification rules: If f0 > f1 and f0 > f2, then the bat is classified as a forager in a C3/C4, 15N-enriched food web. If f1 > f0 and f1 > f2, then the bat is clas-sified as a forager in a C3/C4, 15N-depleted food web. If f2 > f0 and f2 > f1, then the bat is classified as a forager in a C3, 15N-depleted food web.

Resubstitution was used to test the accuracy of Model 3 as described for Model 1 and indicated 100% accuracy within the model dataset.

model application

Our analysis of carbon and nitrogen stable isotope signatures in bat samples from various sites in 10 other states indicates significant differences between many states in terms of carbon and nitrogen signatures (Table I). Each signature shows a continuous distribution over the observed range (Figure 5). All bats were assigned a foraging category using equations 1.1- 3.3 (Table IV; Figure 6).

variation in δ13c acroSS a precipitation gradient

To test whether carbon signatures from bat tissue varied as a function of differences in plant tissue signature along a precipitation gradient, we compared 19 skin tissue samples from six counties across a 330-km gradient from eastern to central Kansas with an annual precipitation gradient of approximately 60 cm. We analyzed the signatures using 1) a regression against thirty-year historical averages of each county (Policy Research Institute, 2004) and 2) an ANOVA with groupings based upon the “precipitation belt” into which the county falls (Spatial Climate Analysis Service, 2004).

figure 4. Plot of δ15N versus δ13C for dataset used for model develop-ment. Hillsborough, NH can be distinguished from Martin and Vigo, IN based on δ13C signatures (Model 1; F2, 39 = 112.92, P < 0.0001) and Vigo, IN can be distinguished from Martin, IN and Hillsborough, NH based upon δ15N signatures (Model 2; F2, 39 = 97.06, P < 0.0001). Each site is indicative of a unique ecotype (Model 3; Wilk’s λ = 0.032, F4, 76 = 87.02, P < 0.0001), and groups used in discriminant analysis do not overlap in each model developed.

f0 = -160.32 + 39.54*δ15N

f1 = -250.92 + 49.45*δ15N

L*δ15N = f1 – f0

f0 = -580.25 – 35.98*δ13C + 50.66*δ15N

f1 = -676.70 – 36.31*δ13C + 60.55*δ15N

f2 = -736.86 – 42.03*δ13C + 52.92*δ15N

Sullivan et al.: Stable iSotope derived habitat-uSe modelS

18

Neither the regression (R2 = 0.16, P = 0.0872) nor the ANOVA (F 2, 16 = 1.46, P = 0.2626) indicates a significant relationship between precipitation and skin tissue carbon signature over the range tested.

DiscussionOur analyses indicate that enrichment of both δ13C

and δ15N from tissues in E. fuscus occurs in the following direction: feces < whole blood < hair < skin < plasma. This

enrichment pattern is in agreement with results obtained by Voigt et al. (2003) in a study conducted on nectar feeding bats that examined carbon signatures of feces, whole blood, hair, and skin. The pattern was consistent at all three sites, although the sites differed in terms of statistical groupings of tissue types (Table III).

Our data suggest the appropriateness of using only one non-invasive sampling protocol (wing biopsies) to infer iso-topic signatures of other tissues. The regression equations generated from highly significant regressions with high cor-relation coefficients (Figure 3; CI [confidence interval] and PI [prediction interval] equations available upon request) may allow other researchers who have retrieved hair, blood, or plasma samples from specimens of the big brown bat to extrapolate skin tissue values and subsequently use the model formulas (1.1- 3.3) to categorize the habitat-use of individual bats or populations.

These interpretations, while supported strongly by our results, invite further experimentation in the form of multi-seasonal and multiple tissue samples in order to be broadly applicable in studies of insectivorous bats, because it is clear that seasonal variation of carbon signatures may exist, par-ticularly as a result of seasonal migration (Fleming, Nuñez & Sternberg, 1993) or seasonal fat deposition or mobiliza-tion (Barboza, Farley & Robbins, 1997).

The lack of consistency in Tukey-Kramer HSD test groupings for sites across tissue types lends further support for this reservation (Table III, although the hair samples may yield unreliable δ13C signatures as they were not chloroform-methanol washed during sample preparation, Voigt et al., 2003). Seasonal variation in hair tissue can be expected if samples taken before and after seasonal molt are compared (Cryan et al., 2004) or if bats feed in habitats from spring through autumn that are alternately dominated by C3 and C4 plants. For example, if bats feed in the Great Plains of North America, which transitions from C3 in the spring to C4 in the summer (Ehleringer, Cerling & Helliker, 1997), then seasonal differences in isotope signatures of diet within a given region can be expected, with hair maintaining a record of diet at the time of molting. Notwithstanding this possibility, our predictive models make it possible to use tissue collected in summer months for making inferences about geographic variation in dietary habits.

The large dataset from disparate ecogeographies (Table I; n = 329) validates the groups we used in model develop-ment. Samples from Hillsborough, NH were used to repre-sent C3 habitat, and the carbon signatures of bats at that site

figure 5. Distributions and box plots (black squares represent outliers) of a) δ13C and b) δ15N signatures showing the continuous but non-normal distributions of both δ13C and δ15N (W = 0.945, P < 0 and W = 0.963, P < 0.0001, respectively). The dashed line bisecting each plot indicates the cutoff point for each univariate model categorization, as in equations 1.1-1.3 (a) and 2.1-2.3 (b). For example, equations 1.1-1.3 (a), bats with δ13C signatures greater than -22.82 ‰ are categorized as foragers in mixed C3/C4 habitat and those with signatures below this cutoff are categorized as foragers in C3 habitat.

table iv. Results of model application pooled across sites. δ13C and δ15N values represent averages of all bats categorized into each group by each model.

Model Habitat type n δ13C δ15N

1 C3/C4 139 -21.11 ± 0.10 C3 190 -23.98 ± 0.08 2 15N enriched 214 10.50 ± 0.07 15N depleted 115 8.27 ± 0.10 3 C3/C4, 15N enriched 156 -21.85 ± 0.12 10.73 ± 0.09 C3/C4, 15N depleted 33 -22.16 ± 0.27 8.15 ± 0.19 C3, 15N depleted 140 -23.92 ± 0.13 8.97 ± 0.09

ÉcoScience, vol. 13 (1), 2006

19

(-24.79 ± 0.13‰) represent the lightest carbon signatures of any area tested, while the sites used as groups indicative of C3/C4 habitat (Martin, IN: -20.84 ± 0.17‰; Vigo, IN: -20.37 ± 0.38‰) are among the heaviest carbon signatures, with only samples from Nebraska being significantly heavier (-18.83 ± 0.20‰). Likewise, the two sites used to repre-sent nitrogen-depleted food chains had the lightest δ15N values of any sites sampled (Martin, IN: 8.06 ± 0.12‰; Hillsborough, NH: 8.11 ± 0.13‰), and the site used in the model to represent the nitrogen-enriched food web (Vigo, IN: 10.12 ± 0.18‰), while not the heaviest, was signifi-cantly lighter than only one other site (Larimer, CO: 11.72 ± 0.32‰). This indicates that the sites we used to develop the models appropriately represent the extremes of carbon and nitrogen signatures, even when compared to a large dataset of samples from diverse geographies.

Our dataset of samples from disparate ecogeographies shows a continuous distribution in terms of both carbon and nitrogen signatures rather than indicating bimodal distribu-tions as would be expected if bats displayed high fidelity to

foraging in either C3 or C4 habitats and nitrogen-enriched or nitrogen-depleted habitats (Figure 5). This lends support for non-specific habitat use wherein big brown bats forage opportunistically in different habitat types.

This interpretation is in agreement with a previous study that describes the success of big brown bats in inhab-iting urban–rural interfaces (Duchamp, Sparks & Whitaker, 2004). In that study, radio-tracked big brown bats roosted almost exclusively in anthropogenic structures, which is in agreement with other studies (Kunz & Reynolds, 2003), and foraged mostly in agricultural, wooded, and urban areas. Additionally, the wide foraging range of non-reproductive bats (mean = 19.03 ± 5.58 km2), when combined with the observation that each bat foraged in different habitat types on subsequent nights (Duchamp, Sparks & Whitaker, 2004), supports our interpretations in the present study. Our attempts to categorize bats therefore indicate a preponderance of for-aging in, rather than a fidelity to, specific habitat types.

Based upon model application to unknown samples (Figure 6), in Kansas, Nebraska, Colorado, and Montana (New Mexico omitted due to small sample size), most bats appear to forage in nitrogen-enriched C3/C4 habitats. Most samples retrieved from New Jersey, Rhode Island, and New York indicated foraging in nitrogen-depleted C3 habitats. Samples retrieved from Kentucky appear to show forag-ing nearly equally in each habitat type used in the model, whereas bats sampled from Wisconsin appear to have for-aged equally in nitrogen-enriched C3/C4 habitats and nitro-gen-depleted C3 habitats. These results and the results of the fecal analysis suggest that big brown bats likely play an important role in pest control in agricultural areas (Kansas, Nebraska, Colorado, Montana, Kentucky, and Wisconsin) as well as in forested and developed areas (New Jersey, Rhode Island, New York, Kentucky, and Wisconsin).

The results from stable isotope analysis in the present study are consistent with those obtained by Des Marais et al. (1980) in a study of guano produced by insectivorous bats at Carlsbad Cavern (Brazilian free-tailed bat, Tararida brasil-iensis, and the cave myotis, Myotis velifer). In their analysis, they found two classes of branched alkanes that were derived from insects (with δ13C averages of –20.7 and –23.2‰) in bat guano. They speculated that the heavier class of alkanes was derived from plants of C4/CAM origin, whereas the lighter class was derived from plants utilizing a C3 pathway. Our results suggest similar differences between pooled aver-ages of skin tissue from bats categorized as having foraged in C3 (-23.98‰) versus C3/C4 (-21.11‰) dominated habitats (Table IV). While this comparison may be compromised because of the greater variability and lighter nature of iso-topes in feces when compared to skin (Figure 3a), the aver-age difference of 2.5‰ between C3 and C4 inputs observed by Des Marais et al. (1980) is remarkably close to the average difference of 2.87‰ when pooled δ13C values of samples categorized as “C3” and “C3/C4” habitat foragers are compared in the present study. These results lend further support for the use of isotopic signatures in assessing forag-ing habits of insectivorous bats.

Our effort to evaluate foraging habits of big brown bats on a finer geographic scale by coupling variation in

figure 6. Result of model application to dataset of samples from unknown ecogeographies by state, as per a) Model 1 (Equations 1.1-1.3), b) Model 2 (equations 2.1-2.3), and c) Model 3 (equations 3.1-3.3). The y-axis indicates the fraction of total samples from each state that were cat-egorized as foragers in each habitat type. Sample sizes are presented below the state names in each panel (key for states: CO, Colorado; KA, Kansas; KY, Kentucky; MT, Montana; NE, Nebraska; NJ, New Jersey; NM, New Mexico; NY, New York; RI, Rhode Island; WI, Wisconsin).

Sullivan et al.: Stable iSotope derived habitat-uSe modelS

20

δ13C isotope signature to a precipitation gradient in central and eastern Kansas indicated that isotope signatures do not vary over a precipitation gradient of 50-110 cm. This is not surprising given the much greater precipitation clines that have been correlated with shifts in carbon signature of plant tissue in other regions (Stewart et al., 1995). Nonetheless, since distribution of C3 and C4 plants is in part a function of aridity (Ehleringer, Cerling & Helliker, 1997; Tieszen et al., 1979) and because water use efficiency (WUE) of plants affects carbon isotope signature, with carbon signatures generally increasing as WUE increases (Farquhar, O’Leary & Berry, 1982; Farquhar & Richards, 1984), geographic carbon signature gradients in plant tissue are generally cor-related with precipitation gradients (Stewart et al., 1995). Thus, future studies should be designed to investigate varia-tion in isotopic ratios over greater precipitation gradients.

model limitationS and future reSearch

Some limitations are implicit in the above models. We assumed that the habitat sites selected for model develop-ment accurately present carbon and nitrogen signatures that are indicative of foraging in the habitat types surveyed. Expanding the model to include a site representative of nitrogen-enriched C3 habitat and sites representative of exclusively C4 habitat would be expected to add measurably to the precision of the model.

Alternate hypotheses may explain the heavier nitro-gen signature in the case of bats sampled from Vigo, IN. Nitrogen stress, seasonal variation, and different trophic states may have accounted for the ~ 2‰ greater average nitrogen signature of bats at Vigo, IN relative to Martin, IN and Hillsborough, NH (Hobson, Alisauskas & Clark, 1993; McCutchan et al., 2003; Voigt & Matt, 2004). However, sampling at the same time of year (August, 2003) at Vigo and Martin, IN virtually eliminates the possibility of sea-sonal variation causing the shift, and the lack of significant difference between the percent consumption of coleopterans and hemipterans, which accounts for greater than 90% of diet at each site, suggests that nitrogen stress and trophic state were likely consistent between these two sites (Table II). However, the possibility exists that differential consumption of insects of the orders Diptera and Neuroptera could have caused the shift in nitrogen signature. If this was the case, then our models were essentially testing for consumption of these or comparable orders of insects.

Mathematical limitations prevent error estimates from being entered into our models. The model formulas 1.1-1.3 show that the model defines the C3 and C3/C4 transition in terms of skin carbon signature above the densest 50% of the carbon signature observations, thus limiting the number of misclassified samples. In contrast, the model formulas 2.1-2.3 show that the model defines the “15N depleted” and “15N enriched” transition within the densest 50% of the nitrogen signature observations, indicating that more misclassified samples will result from this model (Figure 5).

Additionally the broad range of reported C3 (-20 to -32‰) and C4 signatures (-10 to -20‰; Gannes, del Rio & Koch, 1998) limits the precision of the models. Knowing on a finer scale the δ13C values of representative plant tis-sues at each locale or knowing the δ13C signatures of the

direct food source of big brown bats at each site would permit the development of linear mixing models. This would allow more precise predictions to be made for diets of individual bats or for colonies of bats derived from dif-ferent ecoregions.

Because of the challenges associated with direct obser-vation of feeding behaviour of bats and because of large col-onies formed by some species (e.g., Brazilian free-tail bats) and their extensive nightly dispersal patterns, the develop-ment of linear mixing models for different habitats could be a vital tool for developing models of agricultural pest control and for bat conservation management. Reliable linear mix-ing models would make it possible to predict if certain bats, or bats from certain colonies, preferentially disperse to and forage in C4 agricultural lands or C3 forested regions. We are in the process of collecting both plant tissue and insect sam-ples that are part of the food web at one of the sites surveyed in this paper. After further data collection and analysis, we plan to develop a linear mixing model for these particular sites to more accurately predict site-specific foraging behav-iour of Eptesicus fuscus. Similar plans also are being made for assessing the feeding ecology of the Brazilian free-tailed bats in both natural and agro-ecosystems.

Our study indicates that non-invasive sampling tech-niques are suitable for wide-scale geographic analysis of stable isotopes to more clearly define the feeding ecology of insectivorous bats. Additionally, our study indicates that certain non-invasive samples (blood, skin, hair) can be used to extrapolate the signatures of other tissue types, allowing other researchers to extrapolate skin tissue signatures from blood, skin, or hair and use our models to predict ecotypic foraging patterns.

We have shown that stable isotopes can be used to determine foraging behaviour of the big brown bat (E. fus-cus) and have found over wide geographies that this species does not show fidelity to certain habitat types—results that support previous findings. With further study, it may be pos-sible to develop linear mixing models to determine the role of specific food sources in different ecoregions where bats occur. Our findings support previous studies on birds and bats in the use of stable isotope analysis for determining the local movement and migratory patterns of bats as well as other organisms that may be difficult to observe directly. Such results can be important for conservation management of this and other species of insectivorous bats.

AcknowledgementsWe thank the directors and managers of the following state

rabies laboratories who provided bats that were used in this study: Rhode Island, New York, Wisconsin, Colorado, Montana, Kansas, New Mexico, Nebraska, and Kentucky. We also thank the land owners in Hillsborough County, New Hampshire and Martin and Vigo Counties, Indiana, for allowing us to collect bats on their property and S. Roth, E. Rundquist, and T. O’Shea for collecting additional samples from Kansas and Colorado. We thank Qihao Weng and Binging Liang for producing aerial photos of study sites. We wish to thank C. Voigt, A. Finzi, and M. Pencina for helpful comments on this manuscript. This study was funded in part by Boston University’s Center for Ecology and Conservation Biology, Boston University’s Undergraduate Research Opportunities Program (UROP), and a grant from the

ÉcoScience, vol. 13 (1), 2006

21

National Science Foundation, including Research Experiences for Undergraduates Program, to T. H. Kunz and M. D. Sorenson (REU-9988001), from which K. Buscetta was partially supported.

Literature citedAlbert, A. & J. A. Anderson, 1984. On the existence of maximum

likelihood estimates in logistic regression models. Biometrika, 71: 1-10.

Barboza, P. S., S. D. Farley & C. T. Robbins, 1997. Whole-body urea cycling and protein turnover during hyperphagia and dormancy in growing bears (Ursus americanus and U. arctos). Canadian Journal of Zoology, 75: 2129-2136.

Black, H. L., 1972. Differential exploitation of moths by the bats Eptesicus fuscus and Lasiurus cinereus. Journal of Mammalogy, 53: 598-601.

Brigham, R. M., 1991 Flexibility in foraging and roosting behav-iour by the big brown bat (Eptesicus fuscus). Canadian Journal of Zoology, 69: 117-121.

Coplen, T. B., 1996. New guidelines for reporting stable hydro-gen, carbon, and oxygen isotope-ratio data. Geochimica et Cosmochimica Acta, 60: 3359-3360.

Craig, H., 1953. The geochemistry of the stable carbon isotopes. Geochimica et Cosmochimica Acta, 3: 53-92.

Cryan, P. M., M. A. Bogan, R. O. Rye, G. P. Landis & C. L. Kester, 2004. Stable hydrogen isotope analysis of bat hair as evidence for seasonal molt and long distance migration. Journal of Mammalogy, 85: 995-1001.

Dawson, T. E., S. Mambelli, A. H. Plamboeck, P. H. Templer & K. P. Tu, 2002. Stable isotopes in plant ecology. Annual Review of Ecology and Systematics, 33: 507-559.

Denton, T. M., S. Schmidt, C. Critchley & G. R. Stewart, 2001. Natural abundance of stable carbon and nitrogen isotopes in Cannabis sativa reflects growth conditions. Functional Plant Biology, 28: 1005-1012.

Des Marais, D. J., J. M. Mitchell, W. G. Meinschen & J. M. Hayes, 1980. The carbon isotope biogeochemistry of the individual hydrocarbons in bat guano and the ecology of the insectivorous bats in the region of Carlsbad, New Mexico. Geochimica et Cosmochimica Acta, 44: 2075-2086.

Duchamp, J. E., D. W. Sparks & J. O. Whitaker, Jr., 2004. Foraging-habitat selection by bats at an urban–rural interface: Comparison between a successful and less successful species. Canadian Journal of Zoology, 82: 1157-1164.

Ehleringer, J. R., T. E. Cerling & B. R. Helliker, 1997. C4 pho-tosynthesis, atmospheric CO2, and climate. Oecologia, 112: 285-299.

Farquhar, G. D., M. H. O’Leary & J. S. Berry, 1982. On the relationship between carbon isotope discrimination and the intercellular carbon dioxide concentration in leaves. Australian Journal of Plant Physiology, 9: 121-137.

Farquhar, G. D. & R. A. Richards, 1984. Isotopic composition of plant carbon correlates with water-use efficiency of wheat genotypes. Australian Journal of Plant Physiology, 11: 539-522.

Fleming, T. H., 1995. The use of stable isotopes to study the diets of plant-visiting bats. Symposium of the Zoological Society of London, 67: 99-110.

Fleming, T. H., R. A. Nuñez & L. S. L. Sternberg, 1993. Seasonal changes in the diets of migrant and non-migrant nectarivorous bats as revealed by carbon stable isotope analysis. Oecologia, 94: 72-75.

Gannes, L. Z., C. M. del Rio & P. Koch, 1998. Natural abun-dance variations in stable isotopes and their potential uses in animal physiological ecology. Comparative Biochemistry and Physiology A, 119: 725-737.

Handley, L. L. & J. A. Raven, 1992. The use of natural abundance of nitrogen isotopes in plant physiology and ecology. Plant, Cell and Environment, 15: 965-985.

Herrera, L. G., T. H. Fleming & J. S. Findley, 1993. Geographic variation in carbon composition of the pallid bat, Antrozous pallidus, and its dietary implications. Journal of Mammalogy, 74: 601-606.

Herrera, L. G., K. A. Hobson, A. Manzo, D. Estrada, V. Sánchez-Cordero & G. Méndez, 2001a. The role of fruits and insects in the nutrition of frugivorous bats: Evaluating the use of stable isotope models. Biotropica, 33: 520-528.

Herrera, L. G., K. A. Hobson, L. Mirón, N. Ramírez, G. Méndez & V. Sánchez-Cordero, 2001b. Sources of protein in two species of phytophagous bats in a seasonal dry forest: Evidence from stable-isotope analysis. Journal of Mammalogy, 82: 352-361.

Herrera, L. G., E. Guitierrez, K. A. Hobson, B. Altube & W. G. Diaz, 2002. Sources of assimilated protein in five species of New World frugivorous bats. Oecologia, 133: 280-287.

Hobson, K. A., 1999. Tracing origins and migration of wildlife using stable isotopes: A review. Oecologia, 120: 314-326.

Hobson, K. A., R. T. Alisauskas & R. G. Clark, 1993. Stable-nitro-gen isotope enrichment in avian tissues due to fasting and nutri-tional stress: Implications for isotopic analyses of diet. Condor, 95: 388-394.

Hobson, K. A., H. L. Gibbs & M. L. Gloutney, 1997. Preservation of blood and tissue samples for stable-carbon and stable-nitrogen isotope analysis. Canadian Journal of Zoology, 75: 388-394.

Hobson, K. A. & S. G. Sealy, 1991. Marine protein contributions to the diet of Northern Saw-whet Owls on the Queen Charlotte Islands, British Columbia: A stable isotope approach. Auk, 108: 437-440.

Jardine, T. D. & R. A. Cunjak, 2005. Analytical error in stable isotope ecology. Oecologia. DOI: 10.1007/s00442-005-0013-8. [URL] http://springerlink.com/link.asp?id=kr856p3h43270625

Koopman, K. F., 1993. Order Chiroptera. Pages 137-241 in D. E. Wilson & D. M. Reeder (eds.). Mammal Species of the World: A Taxonomic and Geographic Reference. Smithsonian Institution Press, Washington, DC.

Kramer, C. Y., 1956. Extensions of multiple range tests to group means with unequal number of replications. Biometrics, 12: 309-310.

Kunz, T. H., 2004. Foraging habits of North American insec-tivorous bats. Pages 12-25 in R. M. Brigham, E. K. V. Kalko, G. Jones, S. Parsons & H. J. G. A. Limpens (eds.). Bat Echolocation Research: Tools, Techniques, and Analysis. Bat Conservation International, Austin, Texas.

Kunz, T. H. & D. S. Reynolds, 2003. Bat colonies in build-ings. Pages 91-102 in T. J. O’Shea & M. A. Bogan (eds.). Monitoring Trends in Bat Populations of the United States and Territories: Problems and Prospects. US Geological Survey, Biological Sciences Division, Information and Technology Report, Washington, DC.

Kunz, T. H. & J. O. Whitaker, Jr., 1983. An evaluation of fecal analysis for determining food habits of insectivorous bats. Canadian Journal of Zoology, 61: 1317-1321.

Kurta, A. & R. H. Baker, 1990. Eptesicus fuscus. Mammalian Species, 356: 1-10.

Macko, S. A. & N. E. Ostrom, 1994. Pollution studies using stable isotopes. Pages 45-62 in K. Lajtha & R. H. Michener (eds.). Stable Isotopes in Ecology. Blackwell Scientific Publications, Oxford.

Sullivan et al.: Stable iSotope derived habitat-uSe modelS

22

Mariotti, A., 1984. Natural N-15 abundance measurements and atmospheric nitrogen standard calibration. Nature, 311: 251-252.

McCutchan, J. M., W. M. Lewis, C. Kendall & C. C. McGrath, 2003. Variation in trophic shift for stable isotope ratios of car-bon, nitrogen, and sulfur. Oikos, 102: 378-390.

Mizutani, H., D. A. McFarlane & Y. Kabaya, 1992a. Nitrogen and carbon isotope studies of a bat guano core from Eagle Creek Cave, Arizona, USA. Mass Spectrometry, 40: 57-65.

Mizutani, H., D. A. McFarlane & Y. Kabaya, 1992b. Carbon and nitrogen isotopic signatures of bat guanos as a record of past environments. Mass Spectrometry, 40: 67-82.

Peterson, B. J. & B. Fry, 1987. Stable isotopes in ecosystem stud-ies. Annual Review of Ecology and Systematics, 18: 293-320.

Phillips, G. L., 1966. Ecology of the big brown bat (Chiroptera: Vespertiliondae) in northeastern Kansas. American Midland Naturalist, 75: 168-198.

Policy Research Institute, 2004. Kansas Statistical Abstract 2003. Climate. University of Kansas, Lawrence, Kansas.

Post, D. M., 2002. Using stable isotopes to estimate trophic posi-tion: Models, methods, and assumptions. Ecology, 83: 703-718.

Roth, J. D., 2003. Variability in marine resources affects arctic fox population dynamics. Journal of Animal Ecology, 72: 668-676.

Rubenstein, D. R. & K. A. Hobson, 2004. From birds to butter-flies: Animal movement patterns and stable isotopes. Trends in Ecology and Evolution, 19: 256-263.

Shapiro, S. S. & M. B. Wilk, 1965. An analysis of variance test for normality (complete samples). Biometrika, 52: 591-611.

Spatial Climate Analysis Service, 2004. Oregon State University. [URL] http://www.ocs.orst.edu/pub/maps/Precipitation/Total/States/KS/ks.gif (accessed 2 February 2004).

Stewart, G. R., M. H. Turnbull, S. Schmidt & P. D. Erskine, 1995. C natural abundance in plant communities along a rainfall gra-dient: A biological integrator of water availability. Australian Journal of Plant Physiology, 22: 51-55.

Tieszen, L. L., M. M. Senyimba, S. K. Imbamba & J. H. Troughton, 1979. The distribution of C3 and C4 grasses and carbon isotope discrimination along an altitudinal and moisture gradient in Kenya. Oecologia, 37: 337-350.

Tieszen, L. L., T. W. Boutton, K. G. Tesdahl & N. A. Slade, 1983. Fractionation and turnover of stable carbon isotopes in animal tis-sues: Implications for δ13C analysis of diet. Oecologia, 57: 32-37.

Townsend, M. A., D. P. Young & S. A. Macko, 2003. Kansas case study applications of nitrogen-15 natural abundance method for identification of nitrate sources. Journal of Hazardous Substance Research, 4: 1-22.

Voigt, C. C. & F. Matt, 2004. Nitrogen stress causes unpredictable enrichments of 15N in two nectar-feeding bat species. Journal of Experimental Biology, 207: 1741-1748.

Voigt, C. C., F. Matt, R. H. Michener & T. H. Kunz, 2003. Low turnover of carbon isotopes in tissues of two nectar-feeding bat species. Journal of Experimental Biology, 206: 1419-1427.

Whitaker, J. O., Jr., 1993. Bats, beetles, and bugs. More big brown bats mean less agricultural pests. Bats, 11: 23.

Whitaker, J. O., Jr., 1995. Food of the big brown bat, Eptesicus fus-cus, from maternity colonies in Indiana and Illinois. American Midland Naturalist, 134: 346-360.

Whitaker, J. O., Jr., 2004. Prey selection in a temperate zone insec-tivorous bat community. Journal of Mammalogy, 85: 460-469.

Worthington Wilmer, J. & E. Barratt, 1996. A non-lethal method of tissue-sampling for genetic studies in chiropterans. Bat Research News, 37: 1-3.