AN ASSESSMENT OF SURVEY METHODOLOGY, CALLING ACTIVITY, AND HABITAT ASSOCIATIONS OF WOOD FROGS (RANA SYLVATICA) AND BOREAL CHORUS FROGS (PSEUDACRIS MACULATA) IN A TUNDRA BIOME R. Nicholas Mannan, B.S. A Thesis In WILDLIFE SCIENCE Submitted to the Graduate Faculty of Texas Tech University in Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE Gad Perry Chair David E. Andersen Co Chair Clint W. Boal Co Chair Fred Hartmeister Dean of the Graduate School May, 2008 Texas Tech University, Robert N. Mannan, May, 2008 ACKNOWLEDGEMENTS All work was conducted under ACUC permit number 06021-05 and Parks Canada research permit number Wap-2005-518. I thank Parks Canada, Wapusk National Park; Texas Tech University Department of Natural Resources Management; the U.S. Geological Survey, Texas Cooperative Fish and Wildlife Research Unit; and the U.S. Geological Survey, Minnesota Cooperative Fish and Wildlife Research Unit for providing funding and logistical support. The Eastern Prairie Population Canada Goose Committee of the Technical Section of the Mississippi Flyway Council provided logistical support for this project through the Nestor One field research camp, and M. Gillespie (Manitoba Conservation) coordinated camp support. M. Reiter, C. Henneman, G. Lundie, W. Souer, B. Olson, M. Jones, S. Maxson, B. Luebke, M. Miller, B. Nack, T. Bishop, J. Huener, J. Lawrence, B. McCardle, and M. Roell assisted with data collection. Gad Perry, David E. Andersen, and Clint W. Boal provided guidance and demonstrated unending patience. David B. Wester provided integral statistical support. Finally, I thank my family and office mates for their support in all matters. ii Texas Tech University, Robert N. Mannan, May, 2008 TABLE OF CONTENTS ACKNOWLEDGEMENTS ii LIST OF TABLES v LIST OF FIGURES vii CHAPTER I. OVERVIEW 1 II. AN ASSESSMENT OF ANURAN SURVEY METHODOLOGY IN A TUNDRA BIOME III. 5 Abstract 5 Introduction 6 Study Area 8 Methods 9 Results 14 Discussion 16 FACTORS AFFECTING CALLING ACTIVITY OF BOREAL CHORUS FROGS (PSEUDACRIS MACULATA) AND WOOD FROGS (RANA SYLVATICA) NEAR CAPE CHURCHILL, MANITOBA 28 Abstract 28 Introduction 29 Study Area 30 Methods 31 iii Texas Tech University, Robert N. Mannan, May, 2008 IV. Results 33 Discussion 34 HABITAT ASSOCIATIONS OF BOREAL CHORUS FROGS (PSEUDACRIS MACULATA) AND WOOD FROGS (RANA SYLVATICA) IN A TUNDRA BIOME 43 Abstract 43 Introduction 44 Study Area 46 Methods 47 Results 53 Discussion 55 LITERATURE CITED 72 APENDIX A 77 AUTOMATED RECORDER COMPONENTS iv 77 Texas Tech University, Robert N. Mannan, May, 2008 LIST OF TABLES 3.1 Results of a regression analysis of a linear mixed effects model used to model wood frog call activity, near Cape Churchill, Manitoba, Canada, 2007. 3.2 41 Results of a regression analysis of a linear mixed effects model used to model boreal chorus frog call activity, near Cape Churchill, Manitoba, Canada, 2007. 4.1 42 Mean values of habitat variables at detection and non-detection sites for wood frogs and boreal chorus frogs near Cape Churchill, Manitoba, 2006-2007. 4.2 60 Mean values of percent cover, vegetation height, tallest piece of vegetation, pH, and TDS of triangulation locations and corresponding breeding sites near Cape Churchill, Manitoba, 2006-2007. 4.3 61 Number of sites with and without anuran detections that contained evidence of goose herbivory near Cape Churchill, Manitoba, 2006-2007. 62 v Texas Tech University, Robert N. Mannan, May, 2008 4.4 Correlation coefficients and P-values (in parentheses) of measured habitat variables at potential breeding sites of boreal chorus frogs and wood frogs near Cape Churchill, Manitoba 2006-2007 (n = 204). 4.5 63 Eigenvalues of a correlation matrix and the % variation of habitat variables explained in models of habitat associated with anuran locations near Cape Churchill, Manitoba, 2006-2007. 4.6 64 Results of a post-hoc MANOVA test for within-model significance when predicting anuran presence near Cape Churchill, Manitoba, 2006-2007 (n = 202). 4.7 65 Loading scores of the variables within components of habitat models describing anuran locations near Cape Churchill, Manitoba, 2006-2007. 68 vi Texas Tech University, Robert N. Mannan, May, 2008 LIST OF FIGURES 2.1 Study plots near Cape Churchill, Manitoba, where anuran surveys were conducted in 2006 and 2007. 2.2 22 The change in the numbers of Rana sylvatica detected before and after broadcasting advertisement calls near Cape Churchill, Manitoba, 2006-2007. 2.3 23 The change in the number of Pseudacris maculata detected before and after broadcasting advertisement calls near Cape Churchill, Manitoba, 2006-2007. 2.4 24 Difference in the number of Rana sylvatica detections made by observers and using automated recorders near Cape Churchill, Manitoba, 2006-2007. 2.5 25 The discrepancy between Pseudacris maculata detections made by observers and using automated recorders near Cape Churchill, Manitoba, 2006-2007. 26 vii Texas Tech University, Robert N. Mannan, May, 2008 2.6 The percentage of broadcasts detected by automated recorders conducted at 20 m increments along transects beginning at automated recorders and traveling at bearings of 0, 90, 180 and 270 degrees, near Cape Churchill, Manitoba, 2006-2007. 3.1 Study plots near Cape Churchill, Manitoba, where anuran surveys were conducted in 2007. 3.2 27 38 Mean and standard deviation of the number of anurans detected as a function of (A) date, (B) time of day, (C) temperature, and (D) relative humidity near Cape Churchill, Manitoba, Canada, 2007 (□ = Wood frogs, ▲=Boreal chorus frogs). 4.1 39 Study plots near Cape Churchill, Manitoba, where anuran surveys were conducted in 2006 and 2007. 4.2 59 The negative relationship between component one and wood frog presence, depicted by a logistic regression line near Cape Churchill, Manitoba, 2006-2007. viii 66 Texas Tech University, Robert N. Mannan, May, 2008 4.3 The negative relationship between component one and wood frog presence, depicted by a logistic regression line near Cape Churchill, Manitoba, 2006-2007. 4.4 67 Means and 95% confidence intervals of (A) average vegetation height, (B) % cover by sedge and willow, (C) average height of tallest vegetation, (D) TDS, and (E) pH, with respect to evidence of goose herbivory near Cape Churchill, Manitoba, 2006-2007. 5.1 Diagram of automated recorder components near Churchill, Manitoba, 2006-2007. 5.2 78 Photograph of automated recorder components near Cape Churchill, Manitoba, 2006-2007. 5.3 69 79 Control board for automated recorder near Cape Churchill, Manitoba, 2006-2007. 80 ix Texas Tech University, Robert N. Mannan, May, 2008 CHAPTER I OVERVIEW Two species of anurans, the boreal chorus frog (Pseudacris maculata) and wood frog (Rana sylvatica), inhabit the tundra biome near Cape Churchill, Manitoba, and potentially are susceptible to changes in vegetative structure and composition, caused primarily by foraging geese and changes in climate. Both species of frogs have extensive distributions. The wood frog inhabits areas within much of Canada and Alaska, but also exists along the northeastern seaboard of the United States, through the Great Lakes region and into the northern Midwest states (Stebbins 1951, USGS 2002). The boreal chorus frog is the northernmost of the chorus frogs with a range extending from areas in Arizona and New Mexico through the Midwest, north into Alberta, Ontario, Saskatchewan, Manitoba, and parts of the Northwest Territories (Koch and Peterson 1995, USGS 2002). Both species are common within their ranges and much is known about their life histories (e.g., see Heatwole 1961, Berven 1990). However, both wood frogs and boreal chorus frogs show considerable morphological and behavioral variation across their respective ranges (Pettus and Spencer 1964, Berven 1990), and this variation may influence the way they react to changes in climatic and other environmental perturbations. For example, in contrast to other areas in which boreal chorus frogs and wood frogs are found, there are few plants that provide cover or grow very tall in the 1 Texas Tech University, Robert N. Mannan, May, 2008 tundra biome in northern Manitoba. Sedges (Carex aquatilis, C. rupestris, and C. glacialis) and small willows (Salix planifolia, S. herbacea and S. brachycarp) are the only plants that may provide the shelter and structure important to reproduction in frogs. Thus, factors that affect tundra vegetation likely affect anurans inhabiting this region. Two of the factors most likely to affect tundra vegetation along the west coast of the Hudson Bay are changing climate and increased herbivory by geese. Along the west coast of Hudson Bay, geese feed heavily on sedges. Over the past two decades, mid-continent light geese [snow geese (Chen caerulescens) and Ross’s geese (Chen rossii)] populations have increased at an annual rate of 5-7% (Batt 1997, Jefferies and Rockwell 2002). Increased grazing has altered vegetation production as well as the overall vegetative composition in some areas (Cargill and Jefferies 1989, Jefferies and Rockwell 2002). Removal of sedges by foraging geese may impact anurans by altering wetland structure and water quality. Reduction of sedges may remove the necessary cover and tall vegetative structure integral for anurans to complete their life cycle. Also, the removal of sedges may lower pH of the water at potential breeding sites, a hydrological characteristic important to the survival of both wood frogs and chorus frogs (Corn et al. 1989, Wikberg and Mucina 2002). Currently, the effects of goose herbivory on anuran-habitat associations are unknown. To assess anuran population status and monitor changes in anuran populations inhabiting the tundra, efficient survey methodology is necessary. Established survey methods for anurans include road-side surveys, listening stations, repeated surveys at selected breeding areas, and counts of egg masses (Shirose et al. 1997, Crouch and Paton 2 Texas Tech University, Robert N. Mannan, May, 2008 2000, USGS 2001, Weir et al. 2005). However, because most tundra landscapes are relatively isolated and experience extreme climatic conditions, many standard anuran survey methods are not be feasible or effective there. Both the relatively short time period during which surveys can be conducted and the difficulty in reaching and supporting field operations are two factors that complicate surveying anurans in the tundra. Survey efforts in tundra ecosystems might be made more efficient in several ways. The first is to increase the detectability of anuran species. Males of some anuran species increase calling frequency after broadcasts of conspecific advertisement calls (Wells and Greer 1981, Sullivan 1985). A positive response to broadcasted conspecific advertisement calls could increase the efficiency of surveys, thus decreasing the number of surveys necessary to accurately assess presence or absence. No information exists regarding how wood frogs or boreal chorus frogs respond to broadcasted advertisement calls. Another possible approach to increase the effectiveness of anuran surveys is to maximize the area surveyed. There are few roads in most subarctic and arctic tundra landscapes, rendering roadside counts impractical. However, automated audio recorders (hereafter automated recorders) have been used at anuran breeding locations to record call activity (Penman et al. 2005); it may be possible to use a series of automated recorders, placed at potential anuran breeding locations, to detect anuran presence and effectively survey large areas. Finally, high-latitude tundra landscapes experience extensive daylight during short summers, providing a brief annual period for anurans to breed. Environmental 3 Texas Tech University, Robert N. Mannan, May, 2008 variables such as light intensity have been suggested to affect the calling patterns of anurans (Oseen and Wassersug 2002). Effective survey protocols should take advantage of peak periods of calling, but currently it is unclear how weather and time of day affect calling activity of anurans in the subartic and arctic tundra (e.g., see Andersen et al. 2005). The following chapters address the issues discussed above. In Chapter I, I evaluate the possibility of soliciting calls from non-calling anurans by manually broadcasting conspecific advertisement calls as well as the use of automated recorders to increase survey coverage. In Chapter II, I describe factors affecting the calling patterns of anurans in a tundra biome. Finally, in Chapter III, I examine habitat selection and associations of tundra-dwelling anurans with regard to vegetation, water quality, and patches of vegetation affected by goose herbivory. Throughout this thesis terminology remains consistent, but is redefined in each chapter because each chapter is structured to stand alone as an independent manuscript. This allows readers to understand individual chapters without reading the whole thesis. All references are presented in a single section at the end of the manuscript. 4 Texas Tech University, Robert N. Mannan, May, 2008 CHAPTER II AN ASSESSMENT OF ANURAN SURVEY METHODOLOGY IN A TUNDRA BIOME Abstract The tundra biome near Cape Churchill, Manitoba is being influenced by global climate change and herbivory from an increasing population of light geese. These environmental changes may impact anuran populations, although little is known about population trends in anurans in the region. The isolation of the region may render traditional anuran survey methods ineffective. I tested two methods of surveying for two anuran species, boreal chorus frogs (Pseudacris maculata) and wood frogs (Rana sylvatica). I solicited calls from non-calling anurans by manually broadcasting conspecific advertisement calls, and I used automated audio recorders to increase survey coverage. I detected 0.38 additional wood frogs per survey when broadcast calls were employed, compared to surveys without broadcasts. I was unable to detect additional boreal chorus frogs by broadcasting conspecific advertisement calls. Using automated audio recorders, I was able to identify anuran presence in a radius of 100 m, but detections of wood frogs were low compared to manual surveys. I suggest that broadcasts of wood frog advertisement calls be implemented into surveys for wood frogs in the tundra biome, and that additional research is necessary to determine whether solicitation can be used to increase detection of other anuran species. 5 Texas Tech University, Robert N. Mannan, May, 2008 Introduction Surveys are integral to wildlife science and are applied in both research and management in a variety of systems. Because of their wide applications, survey protocols must be appropriate for the system under investigation. Design of survey protocols should incorporate considerations of detection efficiencies as well as other abiotic and biotic constraints within a particular region. Survey protocols should also be: (1) consistent through time; (2) effective; (3) efficient; and if possible (4) comparable to other surveys conducted in the same area or of the same species. Because of the diversity of the applications and conditions under which surveys are conducted, survey protocols are ideally regionally and species-specific. The design of an ideal survey protocol is complicated. Often, particular protocol attributes represent a trade-off between information quality and quantity. For example, increasing survey intensity will likely yield a more accurate estimate of the abundance of the species under investigation, but will also increase the time or funds needed to conduct the survey. Ultimately, increasing survey intensity usually leads to a decrease in the total area surveyed. Limitations of particular survey protocols create a demand for new or improved protocols where the need for “trade offs” is minimized. Increasing species detection probability is one approach to reducing the trade off between quality and quantity. An increase in detection probability reduces the number of survey repetitions needed to yield accurate estimates of species numbers (MacKenzie et al. 2002). One approach to increasing detection probability of cryptic species is inducing a response from the animals being surveyed. Biologists often broadcast avian calls to 6 Texas Tech University, Robert N. Mannan, May, 2008 solicit conspecific calls from individuals that would have otherwise gone undetected (e.g., see Conway and Simon 2003). Another approach to reducing the trade-off between quality and quantity is increasing the efficiency of the surveyor. Automated audio recorders (hereafter automated recorders) may be used to detect any vocal species allowing a single person to collect data across a large area. Also, automated recorders have a good chance of detecting cryptic vocal species that call intermittently, thereby reducing the need for multiple surveys. Audio recorders have been placed at anuran breeding locations to record call activity (e.g., see Penman et al. 2005). However, the use of automated recorders as a surveying method is relatively new and is still being developed. Recent work has identified two species of anurans, the boreal chorus frog (Pseudacris maculata) and wood frog (Rana sylvatica), inhabiting the tundra biome near the coast of the Hudson Bay in Manitoba (e.g., see Boal and Andersen 2003, Andersen et al. 2005, Reiter et al. in review). Both species occupy the tundra across a gradient from coastal tundra through interior sedge meadow/wetland to the tundra-boreal forest interface. The subarctic tundra in northern Manitoba is covered in snow for most of the year. However, for four months during the summer, the tundra becomes a marsh providing nesting grounds for many bird species as well as breeding habitat for the boreal chorus frog and the wood frog (Boal and Andersen 2003, Parks Canada 2007). The tundra offers a unique set of challenges for anuran surveys rendering most traditional methods inappropriate. Traditional survey methods for anurans include roadside surveys, listening stations, repeated surveys at selected breeding areas, and counts of 7 Texas Tech University, Robert N. Mannan, May, 2008 egg masses (Heyer et al. 1994, Crouch and Paton 2000). Much of the tundra in North America is roadless and summer travel is limited to helicopters or distances that can be walked (Parks Canada 2007). This renders roadside surveys impossible and repeated visits to listing stations and egg mass counts, if conducted with helicopters, expensive or, if walked, time intensive. Also, in Wapusk National Park where I conducted the study, a local population of polar bears (Ursus maritimus) makes camping and nighttime surveys unsafe and requires that all daytime work be conducted in pairs, doubling the expense of survey efforts. Currently, no anuran survey protocol has been created for the tundra biome. In 2006 and 2007, I expanded upon previous work (Boal and Andersen 2003, Andersen et al. 2005, Reiter et al. In review) and conducted anuran surveys in Wapusk National Park near Cape Churchill, Manitoba to: (1) evaluate the response of boreal chorus frogs and wood frogs to broadcasted conspecific advertisement calls; and (2) compare the effectiveness of automated recorders and manual surveys for detecting anuran presence in the subarctic tundra biome. Study Area Wapusk National Park is located on the southwest side of Hudson Bay, Manitoba, Canada. The park boundary is 35 km southeast of the town of Churchill, and the park covers approximately 11,475 km2. The study area is located inside the park in the subarctic tundra biome, which encompassed a matrix of small upland ridges and lowland, sedge-dominated marshes and included a mix of semi-permanent and permanent water 8 Texas Tech University, Robert N. Mannan, May, 2008 bodies throughout. Permafrost occurs near the surface, rendering most water bodies very shallow. Winter temperatures are as low as -50 C with an average of -26 C and summer temperatures range from -10 to 35 C with an average of 11 C (Parks Canada 2007). The town of Churchill, located approximately 65 km northwest of Nestor One, the research camp within the park, receives, on average, 436.1 mm of precipitation a year (Environment Canada 2004). Most recreational uses are restricted in the park, except for traditional uses such as hunting, trapping, fishing, and egg-collecting by local residents and First Nations members. I conducted field work primarily in the area surrounding Nestor One, a goose research station located approximately two km from the Hudson Bay coastline south of Cape Churchill (Easting: 0489270, Northing: 6502207; NAD 27). I collected data within two 12.6 km2 study plots extending from near the coast of the Hudson Bay inland. The study plots were situated 8 km apart and each had a diameter of 4 km (Fig. 1.1). All work was conducted under the Texas Tech University Animal Care and Use Committee permit number: 06021-05 and the Parks Canada research permit number: Wap-2005-518. Methods Surveys with broadcasted calls In the summers of 2006 and 2007, I conducted anuran surveys at potential breeding sites to evaluate the response of boreal chorus frogs and wood frogs to broadcasted conspecific advertisement calls. To select potential breeding sites, I created two 12.6 km2 circular study plots: a northern study plot established in 2006 and a 9 Texas Tech University, Robert N. Mannan, May, 2008 southern study plot established in 2007. These study plots were located approximately 4 km from Nestor One to avoid disturbance to goose nests under observation as part of Canada goose (Branta canadensis interior) monitoring activities. Study plots began at the coast of the Hudson Bay and had a diameter of 4 km that extended inland (Fig. 1.1). I used ARCGIS [version 9.1] (ESRI, Redlands, Calif., USA; use of trade names does not imply endorsement by the U.S. Geological Survey, the University of Minnesota, or Texas Tech University) to randomly select 57 locations within the northern plot and 60 locations within the southern plot. Buffer zones ensured locations were not closer than 200 m from each other and the number of survey locations was dictated by the time and logistical support available for survey activities. On the first visit to each random location, I walked to the nearest potential anuran breeding site to establish locations for surveys (described below). Because water bodies with depths ≤10 cm were likely to dry up within two weeks, I defined potential anuran breeding sites as water bodies deeper than 10 cm. Between 30 May and 18 June 2006, I conducted three broadcast surveys (described below) at each breeding site at ~ 6 day intervals. Between 31 May and 11 July 2007, I repeated surveys at 27 of the randomly selected potential breeding sites surveyed in 2006 and conducted surveys at 60 additional potential breeding sites in my southern study plot (Fig. 1.1). I surveyed the 27 potential anuran breeding sites in my northern study area in both 2006 and 2007 to assess annual variation in habitat use by anurans, and to account for potential changes in vegetation between years. Whether individual frogs reuse the same breeding locations between years on my study site is not known. For the 10 Texas Tech University, Robert N. Mannan, May, 2008 purpose of this paper, in my analyses, I assumed that results obtained at the same survey location (n = 27) were independent between years. During each survey, I stood 5 m from the edge of the potential breeding site and allowed one minute for anurans to acclimate to my presence. I then recorded the number and species of anurans detected during a three-minute listening period. Following the listening period, I conducted one of two treatments or a control. Treatments were comprised of one minute of advertisement calls of either a boreal chorus frog or a wood frog. The control was one minute with the absence of any broadcast (hereafter null broadcasts). Following the treatment or control, I conducted another three-minute listening period during which I recorded number and species of all anurans detected. I used a random number generator to select which broadcast type would be used in each survey. In 2006, I obtained recorded anuran calls from the U.S. Geological Survey Amphibian Research and Monitoring Initiative [ARMI; USGS Upper Midwest Environmental Science Center webpage (http://www.umesc.usgs.gov), last accessed 26 August 2006]. During 2007, I recorded and broadcasted calls from local anurans. I broadcasted wood frog and boreal chorus frog advertisement calls with a megaphone rotating in a circle and at an average of 67 (range = 64 – 71) and 78 (range = 71 – 80) dB, respectively. I used sound levels slightly louder than naturally occurring advertisement calls to ensure that broadcast stimuli reached all nearby anurans. All advertisement broadcasts used during this study were comprised of two to three advertising males of either wood frogs or boreal chorus frogs. Anurans sometimes call at the same time, creating call overlap. Some anurans avoid this by calling sequentially (Fellers 1979, 11 Texas Tech University, Robert N. Mannan, May, 2008 Sullivan 1985, Swartz 1987). To minimize the likelihood that broadcasting overlapping calls acted as a call deterrent, the broadcasts I used contained no call overlap. To supplement the number of surveys at potential breeding sites occupied by anurans, in 2007 I also systematically searched both study plots for presence of anurans. I repeatedly walked (>2 times) eight transects located at 1-km intervals spanning study plots. When I detected anurans, I walked to the breeding site from which the anurans were calling and marked the location. I returned to 11 of these locations and conducted < 3 broadcast surveys in the same fashion described above. Comparison between audile surveys and automated recorders In 2006 and 2007, I placed 11 automated recorders at locations I thought likely to contain anurans based on characteristics of vegetation and water body. Recorders were separated from each other by > 200 m, and were programmed to record ambient sounds for three minutes. I chose three minutes because most anuran species are detected within the first 3 minutes of a survey (Shirose et al. 1997). In 2007, after the end of the anuran calling season, I tested the capability of automated recorders to detect broadcasted calls at different distances from four directions. I broadcasted wood frog mating calls at 20 m increments from the recorder out to 160 m along four transects. Transects began at the recorders and traveled away from the recorders at bearings of 0, 90, 180 and 270 degrees. I then reviewed the recordings and identified positive detections at increasing distances. All broadcasts were at an average of 67 (range = 64 – 71) dB and conducted when wind speeds were below 10 km. 12 Texas Tech University, Robert N. Mannan, May, 2008 Statistical analysis To compare the mean number of anurans of each species detected during the three-minute listening period before and after broadcasts I conducted Wilcoxon matchedpairs tests because my anuran count data were not normally distributed (Conover 1999). I also conducted a Wilcoxon matched-pairs test to compare the mean number of anurans of each species detected three minutes before and after null broadcasts. Finally, I conducted a Mann-Whitney U test to compare the mean number of anurans detected during the initial listing period of broadcast surveys and null broadcast surveys. Tests included data only from sites where I knew anurans were present (based on previous detections of calling anurans). I assumed for these tests that breeding sites were independent and frogs did not move among sites between surveys, a common assumption in anuran surveys (Heyer et al. 1994). I also treated surveys at the same site in different years as independent. I only included the first survey of each treatment type from sites in the analyses to avoid pseudoreplication. Anuran calls periodically overlapped, occasionally making an accurate estimate of anuran numbers difficult. In such cases (n = 2), I recorded a range of frogs present (e.g., 6-8). For analyses, I used the lowest number in each range to reduce overestimation. To compare the effectiveness of automated recorders and audile surveys for detecting anuran presence, I conducted 25 simultaneous audile surveys at five sites during a subset of the automated recorded surveys. My count data were not normally distributed, therefore, I conducted Wilcoxon matched-pairs tests to compare the mean number of anurans of each species detected by automated recorders and audile surveys. I 13 Texas Tech University, Robert N. Mannan, May, 2008 report means and standard deviations in the results. I made the assumption that each recorder functioned with the same detection efficiency. This was not an unreasonable assumption, as all recorders were assembled from the same parts at the same time. Results Surveys with broadcasted calls I conducted 432 broadcast surveys at 144 randomly selected sites, and 14 surveys at 11 sites identified while walking transects over the course of the two field seasons. I detected >1 anuran at 47 of the 155 sites surveyed. Broadcast types were randomly selected and sometimes a broadcast type was repeated at a site. To avoid duplication of a broadcast type from the same location, I used 99 of the 155 surveys conducted at sites where anurans were present in my analysis. I conducted 41 broadcast surveys and 24 null broadcast surveys at locations where wood frogs had previously been detected. I conducted 18 broadcast surveys and 16 null broadcast surveys at locations where boreal chorus frogs had previously been detected. There was no difference in the number of wood frogs detected before ( x= 0.541 ± 0.931) and after null broadcasts ( x = 0.625 ± 1.13; z = 0.00, P > 0.999). Likewise, there was not a significant difference in the number of boreal chorus frogs detected before ( x = 1.06 ± 1.80) and after null broadcasts ( x = 1.13 ± 1.74; z = 0.00, P > 0.999). Similarly, there was no apparent difference in detections of boreal chorus frogs pre broadcast ( x = 0.889 ± 1.37) and post broadcast ( x = 1.00 ± 1.53; z = 0.00, P = 0.999; Fig. 1.3). In contrast, wood frogs responded positively to the broadcasting of conspecific 14 Texas Tech University, Robert N. Mannan, May, 2008 advertisement calls. The mean number of wood frogs detected pre broadcast ( x = 0.824 ± 1.38) increased (t = 12.00, z = 2.73, P = 0.006; Fig. 1.2) to 1.24 ± 1.51 post broadcast. There was no difference in the number of wood frogs detected during the initial listing period of broadcast surveys ( x = 0.824 ± 1.38) and null broadcast surveys ( x= 0.541 ±0.931; adjusted z = 0.938, P = 0.348). Similarly, there was no difference in the number of boreal chorus frogs detected during the initial listing period of broadcast surveys ( x = 0.889 ± 1.37) and null broadcast surveys ( x = 1.06 ± 1.80; adjusted z = 0.00, P > 0.999) Comparison between audile surveys and automated recorders I conducted 25 audile surveys simultaneously with automated recorders over two field seasons. Automated recorders ( x = 0.600 ± 0.867) detected significantly fewer wood frogs compared to audile surveys ( x =0.960 ± 1.27, z = 2.20, P = 0.0277; Fig. 1.4). In contrast, there was no difference between the number of chorus frogs detected with automated recorders ( x = 1.72 ± 1.31) compared to audile surveys ( x = 1.44 ± 1.45; z = 3.30, P = 0.0009; Fig. 1.5). I tested the capability of four automated recorders to detect broadcasts. Automated recorders detected 100 percent of all broadcast up to 100 m (Fig 1.6). No automated recorders detected broadcasts at 160 m (Fig 1.6). 15 Texas Tech University, Robert N. Mannan, May, 2008 Discussion The subartic tundra is experiencing habitat changes due to a variety of factors, but changes in plant communities resulting from herbivory by geese have recently received considerable attention (Batt 1997, Jano et al. 1998). The mid-continent light goose (lesser snow geese [Chen caerulescens] and Ross’s geese [Chen rossi]) population has increased 5-7% percent annually over the past 20 years and altered much of the vegetation in the coastal and interior wetlands in the vicinity of Cape Churchill, Manitoba, Canada and western and southern Hudson and James Bays (Batt 1997, Jefferies and Rockwell 2002). Increased grazing has altered vegetation production and the overall vegetative composition in some areas (Cargill and Jefferies 1984, Jefferies and Rockwell 2002). As goose populations grow and the extent of vegetation change increases, it becomes important to monitor local anuran populations and document potential influences of habitat change. Increasing anuran detectability by soliciting calls from non calling anurans and placing automated recorders at anuran breeding locations will aid in the development of appropriate anuran survey methodology in the tundra biome. Surveys with broadcasted calls Anuran calls are important in territoriality, species identification, mate location, and mate choice (Wells 1977, Wells and Greer 1981, Gerhardt 1982, Ryan 1985, Gerhardt et al. 2000). Call overlap is often avoided to maintain the integrity of information transmitted through vocal broadcasts (Fellers 1979, Sullivan 1985, Swartz 1987) and many anuran species avoid calling simultaneously with nearby conspecifics to 16 Texas Tech University, Robert N. Mannan, May, 2008 reduce call overlap. Broadcasting conspecific advertisement calls has been documented to increase call frequency in already calling anurans (Wells and Greer 1981, Benedix and Narins 1999, Gerhardt et al. 2000), but until now, there has been no evidence that broadcasting advertisement calls could increase the number of anurans detected in field surveys. In my study, using broadcasted advertisement calls slightly increased the number of wood frogs detected per survey but did not effectively elicit calls from boreal chorus frogs. Because of the importance of calls and the information within them, it is not surprising that an audile stimulus simulating potential intra-specific competition resulted in a response by one of the species I tested. Which attributes of broadcasted calls cued responses is unknown, and boreal chorus frogs might also respond to a broadcast call if call attributes, such as call volume, call frequency, or pitch, were different from the ones I used. Detecting a change in the number of calling anurans in a field experiment is difficult, primarily because the size of the chorus under observation can affect the precision at which abundances are estimated. In studies conducted in areas with high densities of frogs and large choruses, numbers of calling anurans are sometimes estimated as ranges. For example, the North American Amphibian Monitoring Program (NAAMP) uses call index values (CIV) to estimate numbers of chorusing anurans in three ranges: 1; single calling anuran, 2; multiple calling anurans with no call overlap, 3; multiple calling anurans with call overlap. The number of anurans each CIV value represents is species specific (e.g., wood frogs CIV 2 = 3-5 frogs). Representing numbers 17 Texas Tech University, Robert N. Mannan, May, 2008 of calling anurans in ranges renders rough estimates, which makes detections of small changes in anuran numbers difficult. The density of anurans in the tundra landscape in which I worked is low (Andersen et al. 2005), allowing me to estimate anuran numbers without using ranges. Estimating anuran numbers, as opposed to ranges, likely contributed to my ability to detect an increase of roughly 0.4 wood frogs per survey. My results suggest that the decision of male wood frogs to call is affected by the presence of intra-specific competition. The function of the anuran advertisement calls is thought to be primarily mate attraction and location. Female anurans often choose mates that call more often or with more volume (Ryan 1985). My call broadcasts simulated an increase in local calling activity. Increased calling activity may serve as an indicator of the presence of potential mates, and initiating calling in the presence of potential mates would be advantageous. My results support previous work that suggested that part of the decision of wood frogs to call is based in the perception of intra-specific competition (Wells and Greer 1981, Burmeister et al. 1999). Comparison between audile surveys and automated recorders The estimated number of calling wood frogs detected by automated recorders and audile surveys differed. Automated recorders collapsed the three-dimensional aural world into a single dimension, which may help explain the discrepancy between the numbers of wood frogs detected by audile surveys and automated recorders. The ability to pinpoint the source of calls likely contributed significantly to the ability of a human estimating the number of calling anurans. Wood frogs make a series of single-note chirps resembling the 18 Texas Tech University, Robert N. Mannan, May, 2008 sound of boiling water, but with a higher pitch. During a chorus, wood frogs do not always avoid call overlap. Without pinpointing wood frog locations, call overlap makes it difficult to distinguish among individual frogs on a recording. Although I did not detect a statistical difference in the number of detections made during audile surveys and by automated recorders, more chorus frogs were detected by automated recorders than were detected by observers during five of 25 simultaneous surveys. Boreal chorus frogs usually call with no overlap, or in tandem pairs. In a chorus, if one frog changes frequency others shift frequency to avoid call overlap (Schwartz 1987). Without pinpointing the source of calls it is difficult to distinguish between boreal chorus frogs shifting frequency and new frogs joining the chorus. Shifts in frequency may have acted as false positives, inflating the number of boreal chorus frogs I detected by recorders. Using automated recorders in my study did not result in accurate assessment of the number of calling wood frogs, suggesting that use of automated recorders for acquisition of specific count data of anurans must be done with caution. The automated recorders did reliably detect anurans up to 100 m in any direction. Therefore, it may be possible to use counts acquired from automated recorders as indices and derive and apply species-specific correction factors to estimate counts. However, such a correction would require knowledge or estimation of region- and species-specific bias in counts. Automated recorders would also likely be effective in monitoring sites for presence or absence, which could be used to monitor frog distribution. Also, because the detection probability for anurans is likely <1, repeated visits to a site may be necessary to 19 Texas Tech University, Robert N. Mannan, May, 2008 accurately determine anuran presence or absence (e.g., see MacKenzie et al. 2002). Automated recorders, if left active for an appropriate amount of time, would likely detect all present anuran species. This would alleviate the need for observers to survey a site multiple times. Although automated recorders did not accurately assess exact numbers of calling anurans it is possible to assess overall changes in numbers of anurans calling. This information would be useful in assessing changes in calling activity across seasons and over time as climate changes. Management implications Several studies have documented increased call frequency of anurans following call broadcasts (Wells 1977, Wells and Greer 1981, Gerhardt 1982, Gerhardt et al. 2000). My study demonstrates that it is possible to solicit a vocal response from non-calling anurans of some species in the field. In situations where single-species surveys are to be implemented, broadcasting conspecific mating calls may be a useful tool for increasing detection levels, especially in areas containing low densities of anurans. In this study only 25% of the sites visited contained anurans. An increase in detection levels decreases the necessary number of survey repetitions thereby decreasing the effort required to survey a particular region with no loss to data quality. A decrease in required survey effort would create more economical survey protocols and facilitate studies encompassing larger areas. However, broadcasting should not be implemented in all anuran surveys. It is likely that in large anuran choruses, where a small number of additional anurans calling 20 Texas Tech University, Robert N. Mannan, May, 2008 will not change estimates of abundance, broadcasting advertisement calls may prove ineffective. Furthermore, it is unclear how broadcasting advertisement calls would affect anuran calling patterns during surveys designed to gather information about an array of species. In this study, using automated recorders did not result in accurate assessment of specific numbers of one species of calling anuran. Therefore, automated recorders in subarctic tundra landscapes should be used for count data with caution, but are likely suitable for documenting anuran presence. Automated recorders function without the presence of an observer, which makes them especially useful for documenting anuran presence or absence in isolated regions where repeated visits are logistically difficult. While this study focused on the use of automated recorders for detecting anurans, the use of automated recorders are not limited to anurans. For example, automated recorders may be useful in avian studies. Avian and anuran surveys use similar methods, such as point counts. Use of automated recorders in avian surveys would offer all the same benefits and drawback that use of automated recorders in anuran surveys offer. 21 Texas Tech University, Robert N. Mannan, May, 2008 Figure 2.1. Study plots near Cape Churchill, Manitoba, where anuran surveys were conducted in 2006 and 2007. 22 Texas Tech University, Robert N. Mannan, May, 2008 28 26 24 22 Number of observations 20 18 16 14 12 10 8 6 4 2 0 -4 -3 -2 -1 0 1 2 3 4 The numbers of R. sylvatica detected after broadcasting minus the number of R. sylvatica detected before broadcasting Figure 2.2. The change in the numbers of Rana sylvatica detected before and after broadcasting advertisement calls near Cape Churchill, Manitoba, 2006-2007. 23 Texas Tech University, Robert N. Mannan, May, 2008 Figure 2.3. The change in the number of Pseudacris maculata detected before and after broadcasting advertisement calls near Cape Churchill, Manitoba, 2006-2007. 24 Texas Tech University, Robert N. Mannan, May, 2008 20 18 16 Number of observations 14 12 10 8 6 4 2 0 -4 -3 -2 -1 0 1 2 3 4 Number of R. sylvatica detections made by the observer minus the number of R. sylvatica detections made by the automated recorders Figure 2.4. Difference in the number of Rana sylvatica detections made by observers and using automated recorders near Cape Churchill, Manitoba, 2006-2007. 25 Texas Tech University, Robert N. Mannan, May, 2008 20 18 16 Number of observations 14 12 10 8 6 4 2 0 -4 -3 -2 -1 0 1 2 3 4 Number of P. maculata detections made by the observer minus the number of P. maculata detections made by the automated recorders Figure 2.5. The discrepancy of Pseudacris maculata detections made by observers and using automated recorders near Cape Churchill, Manitoba, 2006-2007. 26 Texas Tech University, Robert N. Mannan, May, 2008 Figure 2.6. The percentage of broadcasts detected by automated recorders conducted at 20 m increments along transects beginning at automated recorders and traveling at bearings of 0, 90, 180, and 270 degrees, near Cape Churchill, Manitoba, 2006-2007. 27 Texas Tech University, Robert N. Mannan, May, 2008 CHAPTER III FACTORS AFFECTING CALLING ACTIVITY OF BOREAL CHORUS FROGS (PSEUDACRIS MACULATA) AND WOOD FROGS (RANA SYLVATICA) NEAR CAPE CHURCHILL, MANITOBA Abstract Little information exists regarding wood frogs (Rana sylvatica) and boreal chorus frogs (Pseudacris maculata) in the tundra biome, where environmental conditions differ from most of the rest of their breeding ranges. Understanding anuran calling patterns is essential to most anuran survey methodology. During the summer of 2007, I placed automated audio recorders at anuran breeding locations and recorded number of calling anurans near Cape Churchill, Manitoba. I used data loggers and automated call recorders to document date, time of day, temperature, and relative humidity. Automated recorders detected wood frogs between 30 May and 2 July 2007, and boreal chorus frogs between 11 June and 5 July 2007. Calling activity of both wood frogs and boreal chorus frogs was influenced by temperature and day of the year (DOY). Calling activity of boreal chorus frogs was also influenced by time of day and relative humidity. Understanding calling patterns with respect to weather patterns will facilitate future monitoring. 28 Texas Tech University, Robert N. Mannan, May, 2008 Introduction Both boreal chorus frogs (Pseudacris maculata) and wood frogs (Rana sylvatica) have extensive distributions across North America that extend into subarctic tundra in northern Manitoba (USGS 2002, Stebbins 1951). Both species are common within their ranges and much is known about their life histories (e.g., see Heatwole 1961, Berven 1990). However, little information exists regarding either of these species in the tundra biome, where environmental conditions differ from most of the rest of their respective breeding ranges (Boal and Andersen 2003, Andersen et al. 2005, Reiter et al. In review). Tall trees, commonly associated with the breeding sites of both species, are lacking in the tundra and northern latitudes have relatively short summers, long day lengths, and cold temperatures, compared to other breeding habitats of wood frogs and boreal chorus frogs. Both species of frogs show considerable morphological and behavioral variability across their respective ranges and populations occurring in tundra biomes may differ behaviorally in response to the extreme conditions present in the tundra (Pettus and Spencer 1964, Berven 1990). One behavior that may change in response to conditions in the tundra is calling activity. In more temperate regions, both wood frogs and boreal chorus frogs are considered explosive breeders (Oseen and Wassersug 2002, USGS 2002) because all breeding takes place over a relatively short time interval. During this short breeding season, environmental factors have little effect on calling activity. For example, in New Brunswick, Canada, temperature and humidity had little effect on the calling activity of wood frogs (Oseen and Wassersug 2002). However, temperature, humidity, and light 29 Texas Tech University, Robert N. Mannan, May, 2008 intensity affect calling patterns of anurans that exhibit a non-explosive breeding strategy (Oseen and Wassersug 2002). Because anurans are poikiothermic, it is possible that temperature would affect calling activity of anurans in the tundra, where summer freezes are not uncommon. No information exists regarding calling activity of boreal chorus frogs or wood frogs in the tundra biome. Many ecological questions about anurans are addressed using information derived from surveys, and most anuran surveys are rooted in detecting calls. Therefore, understanding factors that affect calling activity in tundra anurans should increase the effectiveness of research and monitoring efforts. Herein, I describe diurnal and seasonal calling activity and the effects of temperature, humidity, and time of day on calling activity of boreal chorus frogs and wood frogs in a tundra biome. Study Area Wapusk National Park, Manitoba, Canada is located on the southwest side of Hudson Bay. The park boundary is 35 km southeast of the town of Churchill and the park covers approximately 11,475 km2. The study area is located inside the park in the subarctic tundra biome, and encompassed a matrix of small upland ridges and lowland, sedge-dominated marshes and included a mix of semi-permanent and permanent water bodies throughout. Permafrost occurs near the surface, rendering most water bodies very shallow. Winter temperatures are as low as -50 C with an average of -26 C and summer temperatures range from -10 to 35 C with an average of 11 C (Parks Canada 2007). The town of Churchill, located approximately 65 km northwest of Nestor One, the research 30 Texas Tech University, Robert N. Mannan, May, 2008 camp within the park, receives, on average, 436.1 mm of precipitation a year (Environment Canada 2004). Few recreational uses occur in the park, except for traditional uses such as hunting, trapping, fishing, and egg-collecting by local residents and First Nations members. I conducted field work primarily in the area surrounding Nestor One, a goose research station located approximately 2 km from the Hudson Bay coastline south of Cape Churchill (Easting: 0489270, Northing: 6502207; NAD 27). I collected data within two 12.6 km2 study plots extending from near the coast of the Hudson Bay inland. The study plots were situated 8 km apart and each had a diameter of 4 km (Fig. 2.1). All work was conducted under Texas Tech University Animal Care and Use Committee permit 06021-05 and Parks Canada research permit: Wap-2005-518. Methods Eleven automated recorders were positioned at locations likely to contain anurans based on characteristics of vegetation and water body. All recorders were situated >200 m apart. Each recorder was programmed to record ambient sounds for 3 minutes at 1.5hour intervals. I chose three minutes because most anuran species are detected within the first 3 to 5 minutes of a survey (e.g., see Shirose et al. 1997). I also placed HOBO data loggers (Onset Computer, Pocasset, Mass., H08-030-08; use of trade names does not imply endorsement by the U.S. Geological Survey, the University of Minnesota, or Texas Tech University) at each automated recorder site to record temperature and humidity in concert with audio recordings. 31 Texas Tech University, Robert N. Mannan, May, 2008 Each automated recorder was left active for three days. On the third day, I collected automated recorders in the evening, reviewed recordings, and then replaced recorders the following morning. During review, I listened to each recording and noted the species and number of anuran calls. At the end of the field season, I collected HOBO data loggers and downloaded data on temperature, relative humidity, time, and day of year (DOY) that corresponded with each 3-minute recording interval. In order to extract anuran count data from the loggers, I assumed that the automated recorders accurately depicted anuran calling activity. In Chapter I demonstrated that anuran count data obtained by automated recorders and audile surveys conducted by an observer differed for some species (Mannan et al., unpublished). However, count data obtained from recorders was species specifically biased, suggesting that count data obtained from recorders can be used as an accurate index to calling activity, but not absolute abundance (Mannan et al., unpublished). My study focuses on the effects of weather and seasonal variables on anuran calling activity not exact anuran numbers. Therefore, for the purposes of this study I treated anuran count data obtained from the automated recorders as an index to calling activity. Statistical Analysis My count data were regarded as Poisson-distributed random variables (e.g., see Fitzmaurice et al. 2004). Variability among locations was not of interest in this study and therefore the location effect was included in the model as a random nuisance variable. Because of logistical considerations, data from the automated recorders were not 32 Texas Tech University, Robert N. Mannan, May, 2008 necessarily collected on the same days of the year. The resulting data set, therefore, was analyzed using the methods of longitudinal data analysis (Fitzmaurice et al. 2004). I used a linear mixed effects model to model anuran count data with a log link function, and included location as a random nuisance variable, and day of year, relative humidity, temperature, and time of day as independent variables. Data were analyzed with the GLIMMIX procedure of SAS. Results In 2007, I placed 11 recorders at potential anuran-calling locations, and three of these failed following a severe storm. I obtained data related to diurnal and seasonal anuran calling patterns from the remaining eight recorders. Count data were collected at eight locations from 30 May (DOY=150) to 9 July 2007 (DOY=190). Automated recorders detected wood frogs between 30 May (DOY=150) and 2 July 2007 (DOY=183), and boreal chorus frogs between 11 June (DOY=162) and 5 July (DOY=186) 2007 (Fig. 2.2). Both species exhibited calling patterns consistent with explosive breeding behavior. Peak calling activity for boreal chorus frogs followed peak calling activity for wood frogs by 12 days. Both species called 24 hours a day throughout their breeding season, although peak calling times were afternoon and early morning (i.e., 1200hrs – 0230 hrs; Fig. 2.2). Wood frogs called in temperatures ranging from -2.4 - 29.5 C. calling activity reassembled a normal distribution with lowest calling activity occurring during temperature extremes (Fig. 2.2). Wood frogs called in relative humidity ranging from 0% to 100% (Fig. 2.2). Increased calling activity was skewed towards lower 33 Texas Tech University, Robert N. Mannan, May, 2008 levels of humidity. Boreal chorus frogs called in temperatures ranging from -1.5 - 31.1C, (Fig. 2.2). Calling activity increased as temperature increased. Boreal chorus frogs called in relative humidity ranging from 0% to 100%, with peak calling activity occurring during low levels of relative humidity (Fig. 2.2). Temperature and DOY explained a significant amount of variation within the calling patterns of wood frogs (Table 2.1). Relative humidity, temperature, DOY, and time of day explained a significant amount of variation within the calling patterns of boreal chorus frogs (Table 2.2). Discussion Both wood frogs and boreal chorus frogs in my tundra study site exhibited calling behavior consistent with an explosive breeding strategy. However, unlike other studies conducted on explosive breeding anurans, in my study, colder temperatures were associated with reduced calling activity of both species within the breeding season (e.g., see Oseen and Wassersug 2002). During late May and June, temperatures in northern Manitoba can drop well below 0 C. Anurans are poikiothermic, and therefore cannot remain active in all temperatures. The inability to remain active in low temperatures may explain the association of calling activity of anurans and temperature on my study area. Despite the positive association between wood frog calling activity and temperature, I documented wood frogs calling during intervals when air temperature was below freezing (Fig. 2.2). This may be attributed to a difference between air and water temperature. Wood frogs spend a great deal of time in the water (Conant and Collins 34 Texas Tech University, Robert N. Mannan, May, 2008 1991). Water temperatures may have remained high enough to support anuran activity during intervals when air temperature dropped below freezing. For this study, DOY was designed to account for the effect of seasonal cues such as day length or other unknown factors perceivable by anurans. A day is a fluctuating set of environmental conditions occurring during a time interval. Within this varying set of conditions are known (e.g., temperature) and unknown cues that trigger the start and stop of anuran calling (e.g., see Conant and Collins 1991, Oseen and Wassersug 2002), which should be especially important in explosive breeding anurans. While we cannot identify specific factors causing the stat and stop of wood frog or boreal chorus frog breeding seasons, the importance of DOY representing a seasonal effect on calling activity in my study is consistent with the explosive breeding strategy of both anuran species. One possible confounding factor related to the association between DOY and calling patterns is that the automated recorders detected wood frogs on the first day of recording. Therefore, it is likely that wood frog calling began before I was able to place automated recorders in the field. If recorders were not in place to capture the beginning of the calling season, then the first day of the recording interval would appear to be an exaggerated abrupt beginning to the wood frog calling season. Therefore, some of the variation in wood frog calling activity explained by DOY may have been an artifact of the timing of the recorder placement. However, it is unlikely that calling began long before automated recorders were activated. When recorders were activated there was still snow around anuran breeding ponds. While wood frogs are known to call before the complete thaw of breeding ponds, it is unlikely that calling began long before my arrival 35 Texas Tech University, Robert N. Mannan, May, 2008 (USGS 2002, Stebbins 1951). Also, wood frogs are known to begin breeding abruptly throughout their breeding range and it seems likely that they began abruptly at my study site as well (USGS 2002, Stebbins 1951). Relative humidity and time of day also accounted for variation in the calling pattern of boreal chorus frogs. The lowest detection levels of boreal chorus frogs began around sunrise (0400 hrs) and continued through 1000 hrs (Fig. 2.2). Peak calling occurred in the afternoon (1400 hrs and continued through the evening. These results suggest that light intensity, related to time of day, may influence calling activity of boreal chorus frogs, a common pattern among anurans (e.g., see Bridges and Dorcas 2000, Oseen and Wassersug 2002). Avoidance of predators and thermoregulation have been suggested as ultimate causes of the effect of light intensity on anuran calling activity. Several bird species occurred in my study site during summer that likely prey on frogs, including sandhill cranes (Grus canadensis), and many of those forage more vigorously during morning hours (Iverson et al. 1985), a time when calling generally was suppressed. Thus, predator avoidance seems a plausible explanation for reduced anuran calling during morning hours in northern Manitoba. There was a negative association between the calling activity of boreal chorus frogs and humidity. Peak calling rate was at a relative humidity of about 30%. Peak calling at low humidity levels is unusual for anurans (e.g., see Fogarty and Viella 2001, Stevens et al. 2002). However, high humidity often accompanies precipitation. There are several reasons for the boreal chorus frog to avoid calling during a precipitation event. First, precipitation can interfere with sound transmission negating any reproductive 36 Texas Tech University, Robert N. Mannan, May, 2008 advantage to calling (Henzi et al. 1995). Second, in the tundra, summer precipitation can come in the form of sleet and is often accompanied by large temperature swings. It may be disadvantageous or physiologically impossible to call during such conditions. High levels of humidity may cue a suppression of calling. Management implications Calling activity of anurans, and thus the ability to detect them in surveys, varies across regions and within seasons (Bridges and Dorcas 2000). Surveys of anurans in a given region should, therefore, be timed according to local patterns. My results suggest that breeding seasons of boreal chorus frogs and wood frogs overlapped, which would facilitate monitoring both species simultaneously. Within the breeding season, temperature influenced the calling activity of both species and time of day was related to calling activity of boreal chorus frogs. Thus, conducting surveys during the evening and in warmer temperatures may maximize survey efficiency. However, because the call activity of the two species differed in their response to some climatic variables (e.g., relative humidity), species-specific surveys may result in higher detection levels for individual species. 37 Texas Tech University, Robert N. Mannan, May, 2008 Figure 3.1. Study plots near Cape Churchill, Manitoba, where anuran surveys were conducted in 2007. 38 Texas Tech University, Robert N. Mannan, May, 2008 A. 3.0 Mean number of anurans detected per site 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 30-M ay 4-Jun 9-Jun 15-Jun 20-Jun 25-Jun 1-Jul 6-Jul 1-Jun 7-Jun 12-Jun 17-Jun 23-Jun 28-Jun 3-Jul 9-Jul Date Figure 3.2. Mean and standard deviation of the number of anurans detected as a function of (A) date, (B) time of day, (C) temperature, and (D) relative humidity near Cape Churchill, Manitoba, Canada, 2007 (□ = Wood frogs, ▲=Boreal chorus frogs). 39 Texas Tech University, Robert N. Mannan, May, 2008 B. 3.0 Mean number of anurans detected per site 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 0:00 1:35 3:11 4:47 6:23 7:59 9:35 11:11 12:47 14:23 15:59 17:35 19:11 20:47 22:23 23:59 Time C. 2.2 2.0 Mean number of anurans detected per site 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -4.3 -2.0 0.4 2.8 5.1 7.5 9.9 12.2 14.6 16.9 19.3 21.7 24.0 26.4 28.8 31.1 Temp (C) Figure 3.2. Continued 40 Texas Tech University, Robert N. Mannan, May, 2008 D. 3.0 Mean number of anurans detected per site 2.5 2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -1.6 5.4 12.5 19.5 26.6 33.6 40.6 47.7 54.7 61.8 68.8 75.8 82.9 89.9 97.0 104.0 Relative humidity Figure 3.2. Continued Table 3.1. Results of a regression analysis of a linear mixed effects model used to model wood frog call activity, near Cape Churchill, Manitoba, Canada, 2007, (n= 1383). Covariate Standard Error F P RH (%) 0.00326 0.64 0.4250 Temp (C) 0.01300 12.95 0.0003 Day of Year 0.00703 123.81 <0.0001 Time 0.2193 1.49 0.2219 41 Texas Tech University, Robert N. Mannan, May, 2008 Table 3.2. Results of a regression analysis of a linear mixed effects model used to model boreal chorus frog call activity, near Cape Churchill, Manitoba, Canada, 2007 (n= 1383). Covariate Standard Error F P RH (%) 0.00317 6.59 0.0103 Temp (C) 0.01192 6.72 0.0096 Day of Year 0.00955 0.0011 0.0011 Time 0.2250 0.0214 0.0214 42 Texas Tech University, Robert N. Mannan, May, 2008 CHAPTER IV HABITAT ASSOCIATIONS OF BOREAL CHORUS FROGS (PSEUDACRIS MACULATA) AND WOOD FROGS (RANA SYLVATICA) IN A TUNDRA BIOME Abstract Anuran populations in the subarctic tundra biome in northern Manitoba are being impacted by multiple factors, including vegetation changes resulting from foraging pressure by an increasing population of light geese. During the summers of 2006 and 2007, I surveyed 204 potential anuran breeding locations in two study plots in the tundra biome within Wapusk National Park, Manitoba, Canada to assess anuran habitat associations. I examined habitat selection and associations of wood frogs (Rana sylvatica) and boreal chorus frogs (Pseudacris maculata) with regard to vegetation and water quality. Both wood frogs and boreal chorus frogs selected sites where vegetation was taller and had a higher composition of sedge (Carex spp.) and willow (Salix spp.). Both species also selected sites with relatively low pH and conductivity (TDS). An index of goose herbivory was negatively correlated with vegetation height and cover by sedge and willow and positively correlated with pH and conductivity. Both wood frogs and boreal chorus frogs were found more commonly in sites with less evidence of recent goose herbivory. 43 Texas Tech University, Robert N. Mannan, May, 2008 Introduction Tundra landscapes are being altered by both biological and physical factors. Over the past two decades, mid-continent light geese [snow geese (Chen caerulescens) and Ross’s geese (Chen rossii)] populations in northern Manitoba have increased at an annual rate of 5-7% (Batt 1997, Jefferies and Rockwell 2002). Recently, the effects of this population increase on tundra vegetation have received some attention. Increased grazing has altered vegetation production as well as the overall vegetative composition in some areas (Cargill and Jefferies 1989, Jefferies and Rockwell 2002). As goose populations continue to grow and the extent of vegetation change increases, the potential for impacts on other species in the tundra and their habitats also increases r (e.g., Sammler 2001). Two anuran species, boreal chorus frogs (Pseudacris maculata) and wood frogs (Rana sylvatica), are among species that may be negatively affected by goose herbivory in the tundra biome. The range of wood frogs extends into Canada and Alaska, along the northeastern seaboard in the United States, through the Great Lakes region and into the northern Midwest states (USGS 2002, Stebbins 1951). The boreal chorus frog is the northernmost of the chorus frogs with a range extending from portions of Arizona and New Mexico through the Midwest, north into Alberta, Ontario, Saskatchewan, Manitoba, and parts of the Northwest Territories (Koch and Peterson 1995, USGS 2002). Because both species exist across a wide geographical range, specific habitat relationships in many regions are unknown. However, both species use vegetation for cover and as anchors for egg mass deposition, making attributes of vegetation likely important habitat characteristics (Conant and Collins 1991, Crouch and Paton 2000). 44 Texas Tech University, Robert N. Mannan, May, 2008 In the tundra biome, there are few plants that grow very tall and that would provide potential breeding vegetation for anurans. In tundra landscapes in northern Manitoba, sedges (Carex aquatilis, C. rupestris, and C. glacialis) and small willows (Salix planifolia, S. herbacea, and S. brachycarp) are the only plants that may provide the shelter and structure needed for anuran egg mass deposition. Geese in some of these tundra landscapes feed heavily on sedges, and the reduction and removal of sedges by foraging geese may alter a tundra biome in three notable ways. First, both wood frogs and boreal chorus frogs attach egg masses to aquatic vegetation, and reduction of sedges may remove the cover and tall vegetative structure necessary for anuran egg mass deposition. Second, removal of sedges may increase water pH at potential breeding sites; pH level is a hydrological characteristic important to the survival of both wood frogs and chorus frogs (Corn et al. 1989, Moore and Klerks 1998, Wikberg and Mucina 2002). Third, foraging geese also potentially increase Total Dissolved Solids (TDS) in the water by disturbing the substrate as they feed, or by removing vegetation and, thereby, altering the filtration processes facilitated by wetland vegetation (Noller et al. 1994). Currently, the influences of goose herbivory in tundra landscapes on anuran habitat associations are unknown. In 2006 and 2007, I conducted anuran surveys in a tundra landscape in northern Manitoba to evaluate associations between anuran presence and vegetative structure, water body depth, and environmental conditions [pH and TDS]. I also created an index based on evidence of goose herbivory and evaluated associations between this index and 45 Texas Tech University, Robert N. Mannan, May, 2008 vegetation height, cover, and water quality. Finally I evaluated associations between anuran presence and this index of goose herbivory. Study Area Wapusk National Park is located on the southwest side of Hudson Bay, Manitoba, Canada. The park boundary is 35 km southeast of the town of Churchill and the park covers approximately 11,475 km2. The study area is located inside the park in the subarctic tundra biome, and encompassed a matrix of small upland ridges and lowland, sedge-dominated marshes and included a mix of semi-permanent and permanent water bodies throughout. Permafrost occurs near the surface, rendering most water bodies very shallow. Winter temperatures are as low as -50 C with an average of -26 C and summer temperatures range from -10 to 35 C with an average of 11 C (Parks Canada 2007). The town of Churchill, located approximately 65 km northwest of Nestor One, the research camp out of which I conducted this study within the park, receives, on average, 436.1 mm of precipitation a year (Environment Canada 2004). Few recreational uses occur in the park, except for traditional uses such as hunting, trapping, fishing, and egg-collecting by local residents and First Nations members. I conducted field work primarily in the area surrounding Nestor One, a goose research station located approximately 2 km from the Hudson Bay coastline south of Cape Churchill (Easting: 0489270, Northing: 6502207; NAD 27). I collected data within two 12.6 km2 study plots extending from near the coast of the Hudson Bay inland. The study plots were situated 8 km apart and each had a diameter of 4 km (Fig. 3.1). All work 46 Texas Tech University, Robert N. Mannan, May, 2008 was conducted under Texas Tech University Animal Care and Use Committee permit 06021-05 and the Parks Canada research permit Wap-2005-518. Methods In the summers of 2006 and 2007, I conducted anuran surveys at potential breeding sites to evaluate associations between anuran occurrence and vegetative structure and environmental conditions [pH and TDS] (see below). To select potential breeding sites, I created two 12.6 km2 circular study plots; a northern study plot established in 2006 and a southern study plot established in 2007. These study plots were located approximately 4 km from Nestor One to avoid disturbance to goose nests under observation as part of Canada goose (Branta canadensis interior) monitoring activities. Study plots began at the coast of the Hudson Bay and had a diameter of 4 km that extended inland (Fig. 3.1). I used ARCGIS [version 9.1] (ESRI, Redlands, Calif., USA; use of trade names does not imply endorsement by the U.S. Geological Survey, the University of Minnesota, or Texas Tech University) to randomly select 57 locations within the northern plot and 60 locations within the southern plot. Buffer zones ensured locations were not closer than 200 m from each other and the number of survey locations was dictated by the time and logistical support available for survey activities. On the first visit to each random location, I walked to the nearest potential anuran breeding site to establish locations for surveys (described below). Because water bodies with depths ≤10 cm were likely to dry up within two weeks, I defined potential anuran breeding sites as water bodies deeper than 10 cm. 47 Texas Tech University, Robert N. Mannan, May, 2008 To describe potential breeding sites, I first used a local coordinate system to select two sample locations within each site. To create a local coordinate system, I estimated the width and length of the water body that comprised the breeding site. I then used these estimated distances as the axes of a hypothetical grid, with 1 m cells, that encompassed the entire water body plus 0.5 m beyond the water body’s edge. I used a random number generator to identify two grid intersections, which served as sample locations within each potential breeding site. Some lakes in my study area lacked vegetation beyond the shoreline and I never observed anurans calling from beyond the shoreline of these lakes. Therefore, if the coordinates of a sample location fell in open water within a lake that contained no vegetation beyond 1 m of the shoreline, all habitat variables (described below) were measured from the point on the shoreline closest to the grid intersection. At each sample location, I dropped a 1x1 m quadrat (constructed of rigid polyvinyl chloride tubing) at my feet. In the case of a shoreline location, half of the quadrat was placed on the shore and half in the lake. Within each quadrat, I measured pH, TDS, water depth at two opposite corners of the quadrat, vegetation height at the four corners of the quadrat, and the tallest vegetation within the quadrat. I chose these variables based on their biological importance to anuran habitat selection and/or survival (Pierce et al. 1984, Corn et al. 1989, Moore and Klerks 1998, Anderson et al. 1999, Crouch and Patton 2000, Kopp and Eterovick 2006). For each potential breeding site, I used the two sample locations to calculate an arithmetic mean of pH, TDS, vegetation height, and tallest vegetation, and used these means in statistical analyses. 48 Texas Tech University, Robert N. Mannan, May, 2008 I also took a digital photo from 1.5 m directly above the quadrat to estimate vegetation composition. Using Microsoft® Paint (Microsoft 2002; use of trade names does not imply endorsement by the U.S. Geological Survey, the University of Minnesota, or Texas Tech University), I placed a grid with 100 intersections on each photo. Using Johnson (1987) as a key, I identified vegetation at each grid intersection to genus, and estimated percent cover as the number of times a genus occurred under an intersection divided by the total number of intersections. For each potential breeding site, I used the photos from each sample location (n = 2) to calculate an arithmetic mean vegetation composition by percent cover. At every potential anuran breeding site, I also visually surveyed the entire site for evidence of goose herbivory, and evaluated herbivory based on evidence of goose foraging (Kerbes et al. 1990). I quantified evidence of goose herbivory into two foraging types: “shoot pulling” and “grubbing”. “Shoot pulling” was evidenced by the presence of uprooted sedges with the base of the plant removed. This is caused when geese uproot vegetation, eat the basal tissue, and leave the rest of the plant. Shoot pulling was separated into two groups; recent, or “new shoot pulling” and past, or “old shoot pulling”, based on the color and condition of the uprooted vegetation. I classified uprooted floating vegetation as new shoot pulling, and classified older and decomposing vegetation as old shoot pulling. I similarly assessed grubbing in the substrate. Grubbing is characterized by overturned small chunks of substrate, about the size of a goose bill. These measures of goose foraging were designed to be correlated with grazing pressure. New shoot pulling indicated only recent grazing and thus the smallest amount of grazing pressure. Old shoot 49 Texas Tech University, Robert N. Mannan, May, 2008 pulling represented a history of grazing, thus indicating heavier grazing pressure. Grubbing normally occurs in areas lacking vegetation emergent from the substrate. I assumed that in areas with no vegetation, above-ground vegetation had previously been removed by geese, and therefore grubbing indicated the highest amount of grazing pressure. To document anuran presence or absence at each potential breeding site, I conducted three surveys at ~ 6 day intervals. During each survey, I stood 5 m from the edge of the potential breeding site and allowed one minute for anurans to acclimate to my presence. I then recorded the number and species of anurans detected during a 3-minute listening period. Following the 3-minute listening period I broadcasted anuran mating calls in an attempt to increase detection rates (described below). I conducted all surveys at each site from the same location. When I detected anurans during a survey, I triangulated the position of the first detected anuran when calling continued long enough for observers to estimate a bearing to the frog. Two observers standing ≥5 m apart estimated the direction from which they heard the anuran and then walked in that direction until their paths crossed at the estimated location from which the anuran called (referred to as the triangulation location). I then measured vegetation in a quadrat centered on the triangulation location, as described above. I used the triangulation locations to represent habitat characteristics selected for by calling anurans within each site. In 2007, to supplement information on the characteristics of randomly located potential breeding sites occupied by anurans, I -systematically searched both study plots 50 Texas Tech University, Robert N. Mannan, May, 2008 for presence of anurans. I repeatedly walked (>2 times) eight transects located at 1-km intervals spanning study plots. When I detected anurans, I walked to the breeding site from which the anurans were calling and triangulated an estimated calling location, as described above. I then estimated the same habitat characteristics at the triangulation location and at two random locations as I did for randomly located breeding sites occupied by anurans. I conducted surveys in two study plots (Fig. 3.1) during the course of the study. In 2006, between 30 May and 18 June, I conducted surveys at 57 random locations in Study Plot One, established in 2006 north of Nestor One. In 2007, between 31 May and 11 July, I repeated surveys at 27 randomly selected potential breeding sites surveyed in 2006, and conducted anuran surveys at an additional 60 potential breeding sites in Study Plot Two, established south of Nestor One. I conducted transect surveys (on both study plots) only in 2007. Little empirical evidence exists about the rate at which vegetation is changing in my study site, and although it is possible that anurans present and detected during surveys in 2006 were present at some of the 27 sites resurveyed in 2007, I assumed my surveys between years to be independent in statistical analyses. Statistical Analysis I evaluated anuran habitat associations at two levels; breeding site and within breeding site. At the site level, I used t-tests to compare habitat characteristics of unoccupied and occupied sites. To conduct this comparison, I used only data collected from the two random sample locations of each site. To evaluate within-site associations, 51 Texas Tech University, Robert N. Mannan, May, 2008 from sites occupied by anurans, I used paired t-tests to compare habitat characteristics of triangulation locations to the mean value of habitat characteristics of the two random sample locations. I also used Fisher’s exact tests to further examine relationships between presence of anurans and evidence of goose herbivory. The variables I measured exhibited high levels of inter-correlation. Because my variables were correlated and it is unclear which variables would be appropriate to eliminate from my analysis, I used Principal Components Analyses (PCA) to assess associations (McGarigal et al. 2000). I used PCA (STATISTICA [release 7]; StatSoft, Tulsa, Okla., USA; use of trade names does not imply endorsement by the U.S. Geological Survey, the University of Minnesota, or Texas Tech University) to extract components from my data and identify characteristics associated with breeding sites occupied by anurans. I selected components with eiganevalues above one to test for effect on anuran presence (McGarigal et al. 2000) and conducted a multivariate analysis of variance (MANOVA) to examine the associations between the selected components and the presence of anurans at potential breeding sites. I then used a MANOVA to test for within-model component significance. I used PCA loading scores to evaluate the importance of variables within components. The absolute value of the loading score of each variable within a component indicates how strongly the component is associated with that variable (McGarigal et al. 2000). The higher the absolute value of a loading score, the stronger the association. As the value chosen to warrant investigation or suggest importance to a component is 52 Texas Tech University, Robert N. Mannan, May, 2008 arbitrary, I used 0.5, a common value used to assess variable importance within a component (McGarigal et al. 2000). Results Across both study plots and years, I detected anurans at 36 (25%) of 144 random potential anuran breeding sites. I detected anurans at an additional 60 breeding sites by walking transects through study plots. Using both methods, I detected a combined total of 48 potential breeding sites where I detected only wood frogs, 18 potential breeding sites where I detected only boreal chorus frogs, and 30 potential breeding sites where I detected both species. For both wood frogs and boreal chorus frogs, sites where I detected frogs contained taller vegetation and higher percent cover of vegetation than sites where I did not detect frogs (Table 3.1). Occupied sites also contained lower pH and TDS than unoccupied sites (Table 3.1). In addition, boreal chorus frogs were present at sites with more cover and taller vegetation than wood frogs (Table 3.1). At the scale of location within occupied sites, both anuran species were detected in areas with taller vegetation and a higher percent composition of sedge and willow (Table 3.2). Both species were detected more frequently at areas without evidence of goose herbivory; boreal chorus frogs were negatively associated with all three measures of goose herbivory, whereas wood frogs were significantly negatively associated with sites that had evidence of old shoot pulling or grubbing (Table 3.3). 53 Texas Tech University, Robert N. Mannan, May, 2008 Most habitat variables were highly intercorrelated (Table 3.4). Therefore, I extracted 10 components from the variables percent cover by sedge and willow, pH, TDS, water depth, tallest vegetation, average vegetation height, new shoot pulling, old shoot pulling, grubbing, and year. Based on eigenvalues >1, I selected components one through four to test for effect on anuran presence. Approximately 69 percent of the variation across breeding sites was incorporated in these four components (Table 3.5). A model containing these four components was associated with both wood frogs (MANOVA; Wilks’ Lambda = 0.75; F4,204 = 16.4, P < 0.001) and boreal chorus fogs (MANOVA; Wilks’ Lambda = 0.70; F4,204 = 21.1, P < 0.001). Within-model significance indicated that components one, three, and four were related to wood frog presence and component one was related to boreal chorus frog presence (Table 3.6). Wood frog and boreal chorus frog presence was negatively associated with component one (Fig. 3.2 and Fig. 3.3). Component one was associated with five variables; tallest vegetation received the heaviest loading score followed by average vegetation height, TDS, percent cover by sedge and willow, and pH. Component three was associated only with average vegetation height. Component four was associated with new shoot pulling and with percent cover by sedge and willow; new shoot pulling received the heaviest loading score (Table 3.7). Sites with new and old shoot pulling or grubbing were negatively associated with vegetation height and percent cover by sedge and willow, and positively associated with pH and TDS (Fig. 3.4). When goose herbivory indices were modeled to represent increasing levels of grazing pressure (i.e., no evidence of goose herbivory, presence of 54 Texas Tech University, Robert N. Mannan, May, 2008 only new shoot pulling, presence of new and old shoot pulling, and grubbing), increased grazing pressure was negatively associated with average vegetation height (F3,204 = 13.1, P < 0.001), tallest piece of vegetation (F3,204 = 22.4, P < 0.001), percent cover by sedge and willow (F3,204 = 4.2, P = 0.006), and positively associated with pH (F3,204 = 2.70, P = 0.047) and TDS (F3,204 = 7.19, P < 0.001) (Fig. 3.4). Discussion On my study, boreal chorus frogs and wood frogs were positively associated with percent cover of vegetation and vegetation height, although boreal chorus frogs tended to favor areas with slightly taller vegetation than wood frogs. Both wood frogs and boreal chorus frogs deposit egg masses on aquatic vegetation (Conant and Collins 1991) and aside from sparse willows, sedge is the tallest vegetation offering structure appropriate for egg mass deposition. Other species of chorus frogs also are associated with tall vegetation (Anderson et al. 1999). Wood frogs are described as canopy generalists, breeding in ponds that are both covered and not covered by tree canopy (Werner and Glennemeier 1999), although associations between tree cover and ground cover have not been assessed. In addition to providing appropriate structure for reproduction, higher cover may also reduce predation risk from birds such as arctic terns (Sterna paradisaea) or sandhill cranes (Grus canadensis). More extensive and higher cover may provide better concealment. The effect of pH on survival of anurans is population-dependant (Pierce 1985). In my study area I found negative relationships between anurans and pH, but all pH levels 55 Texas Tech University, Robert N. Mannan, May, 2008 were within the tolerance ranges of wood frogs and other chorus frogs (Pierce et al. 1984, Corn et al. 1989). Wikberg and Mucina (2002) demonstrated that in sedge-dominated aquatic environments, removal of sedge results in an increase in water pH. On my study area, it may be that anurans are associated with high levels of vegetative cover, which may have lower pH, rather than being associated with lower pH itself. Both Canada geese and light geese return annually to the tundra around Cape Churchill to nest. Light goose numbers are increasing (Batt 1997, Jefferies and Rockwell 2002), and large numbers of light geese migrating to more northerly breeding areas forage in the tundra near Cape Churchill (Jefferies and Rockwell 2002). Goose herbivory has had dramatic effects on tundra vegetation near Cape Churchill, and in some areas, has resulted in almost complete removal of vegetation (Jefferies and Rockwell 2002). Currently there are no standard indices for goose herbivory in a tundra biome. The indices to goose herbivory that I used are based on the pattern of impacts on tundra vegetation from goose foraging described in Krebs et al. (1990). Sites with only evidence of new goose herbivory should display the least severe effects of grazing, and have been impacted least by geese. Sites that contained evidence of old shoot pulling likely had a longer history of grazing and the effects of grazing should be more severe. Grubbing occurs in areas with extensive goose foraging activity because grubbing usually occurs in areas with little above-ground vegetation. This categorization of the extent and history of goose herbivory appeared to reasonably depict grazing intensity, as sites with old shoot pulling, and presumably a longer history of grazing, contained shorter vegetation and higher pH values than sites with no evidence of goose herbivory. Sites with evidence of 56 Texas Tech University, Robert N. Mannan, May, 2008 grubbing contained less cover and shorter vegetation than sites with old shoot pulling. Taller vegetation at sites with new shoot pulling may indicate that geese have only recently begun to use these areas, or that grazing pressure has not been high enough to result in measurable effects on vegetation. The patterns of vegetative cover and pH that I observed, suggest that the index I used to quantify goose herbivory reasonably reflects past levels of grazing pressure. Presence of both wood frogs and boreal chorus frogs was negatively related to evidence of goose herbivory. Because vegetation height and cover are important to both frog species, it is not surprising that there were fewer frogs where there was evidence of more extensive herbivory. Prolonged and heavy grazing by geese can significantly alter the structure and composition of the vegetation that anurans select for breeding sites (Jefferies and Rockwell 2002). Specifically, increased foraging by geese may reduce the amount of sedge, leading to an alteration or loss of habitat important for wood frogs and boreal chorus frogs. One surprising result was the relationship between new shoot pulling and wood frog presence suggested by the PCA analysis. However, given the strong negative association between wood frogs and goose herbivory, it is unlikely new shoot pulling attracts wood frogs. Instead, new shoot pulling likely occurs in sites where sedges are still relatively abundant, and these areas may already be occupied by wood frogs when new shoot pulling takes place. I also observed a negative association between anuran presence and higher levels of conductivity (TDS). Reasons for the association are unclear. However, higher 57 Texas Tech University, Robert N. Mannan, May, 2008 conductivity is correlated to increased grazing pressure, which is associated with decreases in extent of vegetative cover. Vegetation in wetlands filters inorganic and organic substances from water. A decrease in vegetation may hinder these processes, increasing inorganic material in the water, and resulting in higher levels of conductivity (TDS). Therefore, it’s likely that the negative association between anuran presence and increasing TDS is an artifact of the loss of vegetation due to goose herbivory. Management implications Vegetation is important to both anuran species present in the tundra landscape in northern Manitoba and elsewhere (Anderson et al. 1999, Werner and Glennemeier 1999). Wood frogs and boreal chorus frogs appear to select for sites with taller vegetation and higher percent vegetative cover. Increased foraging pressure from goose herbivory may negatively impact both species of anuran inhabiting the tundra biome near Cape Churchill, Manitoba. The extent of effects of goose foraging in this landscape, especially in the freshwater sedge-meadow habitats, has not been adequately assessed. However, it is likely that as grazing pressure increases the number of sites suitable for anuran breeding will also decrease.A more through examination of resulting vegetation damage from grazing pressure must be conducted. I also recommend that sticker population management actions be taken regarding light geese. 58 Texas Tech University, Robert N. Mannan, May, 2008 Figure 4.1. Study plots near Cape Churchill, Manitoba, where anuran surveys were conducted in 20067 and 2007. 59 Texas Tech University, Robert N. Mannan, May, 2008 Table 4.1. Mean values of habitat variables at detection and non-detection sites for wood frogs and boreal chorus frogs near Cape Churchill, Manitoba, 2006-2007. (Bolded values indicate statistical difference between detection and non-detection sites at the 0.05 alpha level: t-tests. a Indicates values are significantly different between species at the 0.05 alpha level; t-tests.). Wood Frog Boreal Chorus Frog Variable Wood Frog (n = 78) Non-detection (n = 126) Boreal Chorus Frog (n = 48) Non-detection (n = 156) Percent Composition of Sedge and Willow 74.0 46.3 72.5 52.1 pH 7.6 8.0 7.6 7.9 137.6 178.1 110.2 178.7 36.2 a 24.3 49.4a 22.5 9.0 a 4.7 15.2 a 3.6 TDS (Total Dissolved Solids) Tallest Vegetation (cm) Average Vegetation Height (cm) 60 Texas Tech University, Robert N. Mannan, May, 2008 Table 4.2. Mean values of percent cover, vegetation height, tallest piece of vegetation, pH, and TDS of triangulation locations and corresponding breeding sites near Cape Churchill, Manitoba, 2006-2007. (Bolded numbers indicates statistical difference at the 0.05 alpha level; t-tests.) Wood Frog Boreal chorus frog Variable Triangulation Location (n = 48) Breeding Site (n = 48) Triangulation Location (n = 38) Breeding Site (n = 38) Percent Composition of Sedge and Willow 87.5 77.5 89.2 80.0 pH 7.5 7.5 7.6 7.7 126.0 123.7 116.55 107.2 51.3 34.4 84.4 54.2 14.8 7.7 31.4 17.9 TDS (Total Dissolved Solids) Tallest Vegetation (cm) Average Vegetation Height (cm) 61 Churchill, Manitoba, 2006-2007. No Evidence Evidence of New of New Shoot Shoot Pulling Pulling Wood Frog Detections 54 Fischer’s Fischer’s Fischer’s No Exact Evidence Exact No Exact Evidence Evidence Test Test of Old Test Evidence of Old of New Results Shoot Results of Results Grubbing Shoot Pulling (P(PGrubbing (PPulling value) value) value) 24 14 64 24 54 62 Texas Tech University, Robert N. Mannan, May, 2008 Table 4.3. Number of sites with and without anuran detections that contained evidence of goose herbivory near Cape 0.66 0.0005 0.003 No Wood Frog Detections 85 41 51 75 64 62 Boreal Chorus frog Detections 23 25 6 42 7 62 No Boreal Chorus Frog Detections 116 0.0007 40 0.0006 58 97 <0.0001 81 75 chorus frogs and wood frogs near Cape Churchill Manitoba 2006-2007 (n= 204). Habitat Variables Percent Cover pH 0.09 -0.27 (0.216) (<0.001) 0.17 -0.19 Veg. Height (0.017) (0.007) 0.29 -0.25 Tallest Veg. (<0.001) (<0.001) Water 0.08 -0.05 Depth (0.233) (0.440) 0.11 -0.22 Grubbing (0.117) (0.002) Old Shoot -0.24 0.17 Pulling (0.001) (0.017) New Shoot 0.03 0.03 Pulling (0.722) (0.677) -0.29 0.29 TDS (<0.001) (<0.001) -0.39 pH X (<0.001) Percent X Cover Year TDS -0.34 (<0.001) -0.26 (<0.001) -0.33 (<0.001) -0.05 (0.453) 0.18 (0.008) 0.28 (<0.001) 0.17 (0.013) X New Shoot Pulling 0.07 (0.344) -0.13 (0.067) -0.17 (0.015) -0.05 (0.482) 0.32 (<0.001) 0.22 (0.002) X Old Shoot Pulling -0.35 (<0.001) -0.13 (0.066) -0.17 (0.013) 0.12 (0.088) 0.02 (0.772) X Grubbing 0.26 (<0.001) -0.18 (0.010) -0.26 (<0.001) -0.31 (<0.001) X Water Depth Tallest Veg. Veg. Height -0.44 (<0.001) 0.10 (0.174) 0.12 (0.094) 0.11 (0.133) 0.84 (<0.001) 0.09 0.204 X X X 63 Texas Tech University, Robert N. Mannan, May, 2008 Table 4.4. Correlation coefficients and P-values (in parentheses) of measured habitat variables at potential breeding sites of boreal Texas Tech University, Robert N. Mannan, May, 2008 Table 4.5. Eigenvalues of a correlation matrix and the % variation of habitat variables explained in models of habitat associated with anuran locations near Cape Churchill, Manitoba, 2006-2007. Component Eigenvalue 1 2.75 % Total Variation Explained 28.5 2 1.86 18.6 46.1 3 1.18 11.6 57.9 4 1.15 11.5 69.4 64 Cumulative % Variation Explained 28.5 Texas Tech University, Robert N. Mannan, May, 2008 Table 4.6. Results of a post-hoc MANOVA test for within-model significance when predicting anuran presence near Cape Churchill, Manitoba, 2006-2007 (n = 202). Wood Frog Intercept Component 1 Component 2 Component 3 Component 4 Degrees of F Freedom 1 163.87 1 41.09 1 0.07 1 6.67 1 18.87 Boreal Chorus Frog P <0.0001 Intercept Component <0.0001 1 Component 0.785923 2 Component 0.0109 3 Component <0.0001 4 65 Degrees of F Freedom 1 87.21 P <0.0001 <0.0001 1 80.53 1 0.319 0.667791 1 0.00 0.241743 1 2.57 0.270363 Texas Tech University, Robert N. Mannan, May, 2008 Wood frog presence 1 0 -4 -3 -2 -1 0 1 2 3 Component one Figure 4.2. The relationship between component one and wood frog detection, depicted by a logistic regression line near Cape Churchill, Manitoba, 2006-2007. 66 Texas Tech University, Robert N. Mannan, May, 2008 Boreal Chorus frog presence 1 0 -4 -3 -2 -1 0 1 2 3 Component one Figure 4.3. The relationship between component one and wood frog presence, depicted by a logistic regression line near Cape Churchill, Manitoba, 2006-2007. 67 Texas Tech University, Robert N. Mannan, May, 2008 Table 4.7. Loading scores of variables within Principal Components of habitat models describing anuran locations near Cape Churchill, Manitoba, 2006-2007.Bolded numbers indicate loading scores of absolute value >0.5. Variable Component 1 Component 2 Component 3 Component 4 Percent Sedge and Willow -0.558303 0.028580 0.287488 0.503835 pH 0.545991 -0.187277 -0.213056 -0.458068 TDS 0.649573 -0.179072 -0.215902 0.019466 0.317061 0.244211 -0.360346 0.702940 0.461116 -0.392509 -0.336038 0.295282 0.401067 0.595045 -0.338211 0.095668 -0.124618 -0.711388 0.138510 0.276933 -0.796087 -0.178313 -0.493451 -0.070232 -0.708621 -0.171146 -0.616651 -0.116700 -0.311491 0.810436 -0.030130 -0.055378 New Shoot Pulling Old Shoot Pulling Grubbing Average Water Depth Tallest Piece of Vegetation Average vegetation Height Year 68 Texas Tech University, Robert N. Mannan, May, 2008 Figure 4.4. Means and 95% confidence intervals of (A) average vegetation height, (B) % cover by sedge and willow, (C) average height of tallest vegetation, (D) TDS, and (E) pH, with respect to evidence of goose herbivory near Cape Churchill, Manitoba, 20062007. 69 Texas Tech University, Robert N. Mannan, May, 2008 Figure 4.4. Continued. 70 Texas Tech University, Robert N. Mannan, May, 2008 Figure 4.4. Continued. 71 Texas Tech University, Robert N. Mannan, May, 2008 LITERATURE CITED Anderson, A.M., D.A. Haukos, and J.T. Anderson. 1999. Habitat use by anurans emerging and breeding in playa wetlands. Wildlife Society Bulletin 27(3):759769. Andersen, D.E., C.W. Boal, and M.E. Reiter. 2005. Wood frog and boreal chorus frog distribution and habitat associations in Wapusk National Park, Cape Churchill, Manitoba: 2005 summary report. Minnesota Cooperative Fish and Wildlife Research Unit, St. Paul, Minnesota, USA. 20pp. Batt, B.D.J., editor. 1997. Arctic ecosystems in peril: report of the Arctic Goose Habitat Working Group. Arctic Goose Joint Venture Special Publication. United States Fish and Wildlife Service, Washington, D.C. and Canadian Wildlife Service, Ottawa, Ontario, Canada. Benedix, J.H. Jr. and P.M. Narins. 1999. Competitive calling behavior by male treefrogs, Eleutherodactylus coqui (Anura: Leptodactylidae). Copeia 1999(4):1118-1112. Berven, K. 1990. Factors affecting population fluctuations in larval and adult stages of the wood frog (Rana Sylvatica). Ecology 71(10):1599-1608. Boal, C.W. and D.E. Andersen. 2003. Pilot study of boreal chorus frog and wood frog distribution and aquatic habitat conditions at Cape Churchill, Manitoba. Unpublished report to Wapusk National Park and the Mississippi Flyway Council. Minnesota Cooperative Fish and Wildlife Research Unit, St. Paul, Minnesota, USA. Bridges, A.S. and M.E. Dorcas. 2000. Temporal variation of anuran calling behavior: implications for surveys and monitoring programs. Copeia 2000(2):587-592. Burmeister, S., W. Wilczynski, and M J. Ryan. 1999. Temporal call change and prior experience affect graded signaling in the cricket frog. Animal Behaviour 57(3):611-618. Cargill, S.M. and R.L. Jefferies. 1984. The effects of grazing by lesser snow geese on the vegetation of a sub-arctic salt marsh. Journal of Applied Ecology 21(2):669-686. Conant, R. and J.T. Collins. 1991. A field guide to reptiles and amphibians: eastern and central North America. Houghton Mifflin Company. Boston, Massachusetts, USA. 72 Texas Tech University, Robert N. Mannan, May, 2008 Conover, W.J. 1999. Practical nonparametric statistics. Third edition. John Wiley and Sons, New York, New York, USA. Conway, C.J. and J.C. Simon. 2003. Comparison of detection probability associated with burrowing owl survey methods. Journal of Wildlife Management 67(3):501511. Corn, P.S., W. Stolzenburg, and R.B. Bury. 1989. Acid precipitation studies in Colorado and Wyoming: interim report of surveys of montane amphibians and water chemistry. Biological Report 80, U.S. Fish and Wildlife Service, Ft. Collins, Colorado, USA. Crouch, W.B. and P.W.C. Paton. 2000. Using egg mass counts to monitor wood frog populations. Wildlife Society Bulletin 28(4):895-901. Environment Canada. 2004. Weather history. <http://www.weatheroffice.gc.ca> Accessed November 2007. Fellers, G.M. 1979. Aggression, territoriality, and mating behavior in North American tree frogs. Animal Behavior 27(1):107-119. Fitzmaurice, G.M., N.M. Laird, and J.H. Ware. 2004. Applied longitudinal analysis. John Wiley and Sons, Hoboken, New Jersey. Fogarty, J.H. and F.J. Viella. 2001. Evaluating methodologies to survey Eleutherodactylus frogs in montane forests of Puerto Rico. Wildlife Society Bulletin 29(3):948-955. Gerhardt, H.C. 1982. Sound pattern recognition in some North American tree frogs (Anuran: Hylidae). Implications for mate choice. American Zoologist 22(3):581595. __________, S.D. Tanner, C.M. Corrigan, and H.C. Walton. 2000. Female preference functions based on call duration in the gray tree frog (Hyla versicolor). Behavioral Ecology 11(6):663-669. Heatwole, H. 1961. Habitat selection and activity of the wood frog, Rana sylvatica, Le Conte. American Midland Naturalist 66(2):301-313. Henzi, S.P., M.L. Dyson, S.E. Piper, N.E. Passmore, and P. Bishop. 1995. Chorus attendance by male and female painted reed frogs (Hyperolius marmoratus): environmental factors and selection pressures. Functional Ecology 9(3):485-491. 73 Texas Tech University, Robert N. Mannan, May, 2008 Heyer, W.R., M.A. Donnelly, R.W. McDiarmid, L.C. Hayek, and M.S. Foster. 1994. Measuring and monitoring biological diversity: standard methods for amphibians. Smithsonian Institution Press, Washington, D.C., USA. Iverson, G.C., P.A. Vohs, T.C. Tacha. 1985. Habitat use by sandhill cranes wintering in western Texas. Journal of Wildlife Management 49(4):1074-1082. ___________ and R.F. Rockwell. 2002. Foraging geese, vegetation loss and soil degradation in an artic salt marsh. Applied Vegetation Science 5(1):7-16. Johnson, K.L. 1987. Wildflowers of Churchill and the Hudson Bay Region. Manitoba Museum of Man and Nature. Winnipeg, Manitoba, Canada. Jano, A.P., R.L. Jefferies, and R.F. Rockwell. 1998. The detection of vegetation change by multitemporal analysis of LANDSAT data: the effects of goose foraging. The Journal of Ecology 86(1):93-99. Koch, E.D. and C.R. Peterson. 1995. Amphibians and reptiles of Yellowstone and Grand Teton National Parks. University of Utah Press. Salt Lake City, Utah, USA. Kopp, K. and P.C. Eterovick. 2006. Factors influencing spatial and temporal structure of frog assemblages at ponds in southeastern Brazil. Journal of Natural History 40(1):29-31. Kerbes, R.H., P.M. Kotanen, and R.L. Jefferies. 1990. Destruction of wetland habitats by lesser snow geese: a keystone species on the west coast of the Hudson Bay. Journal of Applied Ecology 27(1):242-258. MacKenzie, D.I., J.D. Nichols, G.B. Lachman, S. Droege, J.A. Royle, and C.A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83(8):2248-2255. McGarigal, K., S. Cushman, and S. Stafford. 2000. Multivariate statistics for wildlife and ecology research. Springer-Verlag. New York, New York, USA. Microsoft. 2002. Microsoft Windows XP professional version 2002 service pack 2. Redmond, Wasington, USA. Moore, M.K. and P.L. Klerks. 1998. Interactive effect of high temperature and low pH on sodium flux in tadpoles. Journal of Herpetology 32(4):588-592. 74 Texas Tech University, Robert N. Mannan, May, 2008 Noller, B.N., Woods, P.H., and Ross, B.J. 1994. Case studies of wetland filtration of mine waste water in constructed and naturally occurring systems in northern Australia. Water Science Technology 29(1):257-265. Oseen, K.L. and R J. Wassersug. 2002. Enviornmental factors influencing calling in sympatric anurans. Behavioral Ecology 133(4):616-625. Parks Canada. 2007. Wapusk National Park. weather. < http://www.pc.gc.ca/pnnp/mb/wapusk/visit/visit4_e.asp>. Accessed October 2007. Penman, T.D., F.L. Lemckert, and M.J. Mahony. 2005. A cost-benefit analysis of automated call recorders. Applied Herpetology 2(5):389-400. Pettus, D.A and W. Spencer. 1964. Size and metabolic differences in Pseudacris triseriata (anura) from different elevations. Southwestern Naturalist 9(1):20-26. Pierce, B.A. 1985. Acid tolerance in amphibians. Bioscience. 35(4):239-243. __________, J.B. Hoskins, and E. Epstein. 1984. Acid tolerance in Connecticut wood frogs (Rana sylvatica). Journal of Herpetology18(2):159-167. Reiter, M.E., D.E. Andersen, and C.W. Boal. In Review. Anurans in a subarctic landscape near Cape Churchill, Manitoba. Canadian Field-Naturalist. Ryan, M. 1985. The túngara frog. The University of Chicago Press, Chicago, Illinois, USA. Sammler, J.E. 2001. Population trends of tundra-nesting birds of Churchill Manitoba: potential effects of increasing lesser snow goose (Chen caerulescens caerulescens) populations. Thesis. University of Minnesota, St. Paul, Minnesota, USA. Shirose, L.J., C.A. Bishop, D.M. Green, C.J. MacDonald, R.J. Brooks, and N.J. Helferty. 1997. Validation tests of an amphibian call count survey technique in Ontario, Canada. Herpetologica 53(3):312–320. Stebbins, R.C. 1951. Amphibians of western North America. University of California Press. Berkley and Los Angeles, California, USA. Stevens, C. E., A.W. Dimond, and T.S. Gabor. 2002. Anuran call surveys on small wetlands in Prince Edwards Island, Canada restored by dredging of sediments. Wetlands 22(1):90-99. 75 Texas Tech University, Robert N. Mannan, May, 2008 Sullivan, K. 1985. Male calling behavior in response to playback of conspecific advertisement calls in two bufonids. Journal of Herpetology 19(1):78-83. Swartz, J.J. 1987. The function of call alternation in anuran amphibians: a test of three hypotheses. Evolution 41(3):461-471. Upper Midwest Environmental Center. 2006. Upper Midwest Environmental Science Center. (http://www.umesc.usgs.gov/). Accessed 26 August 2006. USGS. 2001. United States Geological Survey. North American Amphibian Monitoring Program NAAMP. (http://www.mp2-pwrc.usgs.gov/naamp/). Accessed 15 October 2005. __________. 2002. United States Geological Survey. Northern Prairie Wildlife Research Center. Chorus Frog. (http://www.npwrc.usgs.gov/narcam/). Accessed 15 October 2005. Weir, L.A., J.A. Royle, P. Nanjappa, R.E. Jung. 2005. Modeling anuran detection and site occupancy on North American Amphibian Monitoring Program (NAAMP) routes in Maryland. Journal of Herpetology 39(4):627-639. Wells, K.D. 1977. The social behavior of anuran amphibians. Animal Behavior 25(3):666-693. __________ and B.J. Greer. 1981.Vocal responses to conspecific calls in a neotropical Hylid Frog, Hyla ebraccata. Copeia 1981(3):615-624. Werner E.E. and K.S. Glennemeier. 1999. Influence of forest canopy cover on the breeding pond distributions of several amphibian species. Copia 1999(1):1-12. Wikberg, S. and L. Mucina. 2002. Spatial variation in vegetation and abiotic factors related to the occurrence of ring-forming sedge. Journal of Vegetation Science 13(5):677-684. 76 Texas Tech University, Robert N. Mannan, May, 2008 APPENDIX A AUTOMATED RECORDER COMPONENTS Components in the automated loggers I used included a digital voice recorder (Olympus vn-2000), digital timer (American Deer Hunter 30556), microphone (Olympus me-15), 12-volt battery (Panasonic LC-R121R3P), and a solar panel (Sun Force 77796). (Fig. 4.1 and Fig. 4.2) All components except the solar panel and microphone were housed inside of a metal box. The microphone protruded from the box through a hole. A chicken wire frame encompassed the entire the box, to reduce mammalian disturbance. The solar panel was fixed to a chicken wire frame. Circuitry was constructed at Texas Tech University (Fig. 4.3). Automated recorders were originally programmed to broadcast advertisement calls after the 3-minute recording interval and then conduct another 3-minute recording interval. However, the speakers failed to function properly and automated recorder broadcasts were never conducted in the field. 77 Texas Tech University, Robert N. Mannan, May, 2008 Figure 5.1. Diagram of automated recorder components near Churchill, Manitoba, 20062007. 78 Texas Tech University, Robert N. Mannan, May, 2008 Figure 5.2. Photograph of automated recorder components near Cape Churchill, Manitoba, 2006-2007. 79 Texas Tech University, Robert N. Mannan, May, 2008 Figure 5.3. Control board for automated recorder near Cape Churchill, Manitoba, 20062007. 80 Texas Tech University, Robert N. Mannan, May, 2008 PERMISSION TO COPY In presenting this thesis in partial fulfillment of the requirements for a master’s degree at Texas Tech University or Texas Tech University Health Sciences Center, I agree that the Library and my major department shall make it freely available for research purposes. Permission to copy this thesis for scholarly purposes may be granted by the Director of the Library or my major professor. It is understood that any copying or publication of this thesis for financial gain shall not be allowed without my further written permission and that any user may be liable for copyright infringement. Agree (Permission is granted.) Robert_ Mannan________________________________ Student Signature __4/22/08_______ Date Disagree (Permission is not granted.) _______________________________________________ Student Signature _________________ Date