HABITAT ASSOCIATIONS OF WOOD FROGS (RANA SYLVATICA), AND EFFECTS OF FRAGMENTATION, IN BOREAL MIXEDWOOD FORESTS by ELKE WIND B.E.S., University of Waterloo, 1992 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Department of Forest Sciences) We accept this thesis as conforming to the required standard UNIVERSITY OF BRITISH COLUMBIA DECEMBER, 1996 ©ElkeWind, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced i degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of The University of British Columbia Vancouver, Canada Date DE-6 (2/88) ABSTRACT I studied the habitat associations of wood frogs (Rana sylvatica) and fragmentation in boreal mixedwood forests of north-central Alberta. I sampled wood frogs, measured habitat characteristics, and assessed the general wetness of 10 sites from 1993 to 1995. The 10 sampling sites were in continuous forest in 1993, but through forest harvesting, six of these sites were made isolated fragments surrounded by 200-m wide clear cuts in 1994. The abundance of wood frogs decreased and average body length increased within 10-ha fragments two years post-harvesting compared to 100-ha and control sites. Forest fragmentation reduces the number of wood frogs, especially impacting small individuals, but the presence of wet patches within a site may mitigate the effects. Analyses of habitat indicate that wood frogs are associated with some characteristics of boreal mixedwood forests, perhaps most closely related to coarse woody material (CWM) approximately 5<11 cm in diameter. CWM of this size class is correlated with trembling aspen {Populus tremuloides) that is 23<38 cm diameter at breast height (dbh). I recommend further research into the effects of timber harvesting on migration and dispersal of amphibians, and for determining habitat requirements of amphibians at stand and landscape levels in boreal mixedwood forests. ii TABLE OF CONTENTS Abstract i i Table of Contents iii List of Tables v List of Figures vi Acknowledgments vii Chapter I. General Introduction 1 Chapter II. Habitat associations of wood frogs (Rana sylvatica) in boreal mixedwood forests Introduction 5 Research Area and Sampling Sites 7 Methods 10 Amphibians 10 Habitat and Vegetation 15 Wet Patches 16 Data Analyses 18 Results 20 Capture Rates 20 Habitat and Wet Patch Descriptions for Sampling Sites 22 Habitat Associations of Wood Frogs 26 Wet Patches and Abundance of Wood Frogs 32 Discussion 32 iii Chapter III. Effect of fragmentation on wood frogs (Rana sylvatica) in boreal mixedwood forests Introduction 40 Methods 42 Amphibians 42 Data Analyses 42 Results 43 Capture Rates 43 Treatment Effects 46 Discussion 52 Chapter IV Conclusions and Recommendations 59 Literature Cited 63 Appendix 1 69 iv LIST OF TABLES Table Page 2.1 Site descriptions for 10 amphibian sampling sites at Calling Lake, 9 Alberta. 2.2 Wood frog capture rates in control sites from 1993 to 1995 at 21 Calling Lake, Alberta. 2.3 Forward stepwise multiple regression (SMR) of total number of wood 27 frogs caught between 1993 and 1995 in 16 vegetation plots within control sites. 2.4 Multivariate discriminant function analysis (MDA) for all frogs captured 28 during the 1993 pre-treatment year, with sites classified as present or absent. Variables used for discrimination were those selected from a stepwise multiple regression (SMR) with all size classes of frogs combined within control sites from 1993 to 1995. 2.5 Forward stepwise multiple regression (SMR) of 'adult' wood frogs caught 30 between 1993 and 1995 in 16 vegetation plots within control sites. 2.6 Multivariate discriminant function analysis (MDA) for all frogs captured 31 during the 1993 pre-treatment year, with sites classified as present or absent. Variables used for discrimination were those selected from a stepwise multiple regression (SMR) with the 'adult' wood frogs within control sites from 1993 to 1995. 2.7 Significant Pearson correlation analysis results between the wetness of the 33 site and wood frog abundance in 1995. 3.1 Wood frog (Rana sylvatica) capture rates for forest fragments and control 44 sites, between 1993 and 1995 at Calling Lake, Alberta. 3.2 Repeated measures analysis of variance (RM-ANOVA) comparing the 47 change in capture rate of wood frogs ('combined' data set) between the first and second year after harvesting among different treatments. 3.3 Repeated measures analysis of variance (RM-ANOVA) comparing the 49 change in capture rate of wood frogs ('adult' data set) between the first and second year after harvesting among different treatments. 3.4 Repeated measures analysis of variance (RM-ANOVA) comparing the 51 change in mean body length of wood frogs between the first and second year after harvesting among different treatments. v LIST OF FIGURES Figure Page 2.1 Layout of amphibian sampling sites, with trapping grid, four vegetation 11 plots, and wet patch search area. 2.2 Body length of wood frogs (snout to end of urostyle) caught between 14 1993 and 1995. 2.3 Tree and shrub species richness (i.e., number of different species of each) 23 within each vegetation plot in control sites. 2.4 Number of stems of dominant tree species (all size classes combined) 24 within four vegetation plots in control sites. 2.5 Percent wetness and number of wet patches within each sampling site 25 in 1994 and 1995. 3.1 Capture rate of wood frogs by site and treatment at Calling Lake, 45 Alberta, 1993-1995. 3.2 Change in capture rate of wood frogs ('combined' data set) among 48 treatments during first and second year after harvesting compared to preharvest capture rate. Bars are standard errors of the means. 3.3 Change in capture rate of 'adult' wood frogs among fragmentation 50 treatments during first and second year after harvesting, compared to preharvest capture rate. Bars are standard errors of the means. 3.4 Change in mean body length of wood frogs (all sizes combined) among 53 fragmentation treatments first and second year after harvesting, compared to preharvest mean length. Bars are standard errors of the means. 3.5 Proportion of frogs caught within different length classes among 54 treatments between 1993 and 1995. vi ACKNOWLEDGMENTS I thank both Dr. Ann Chan-McLeod and Dr. Fred Bunnell for supervising me these past years. I appreciate the time and effort that they have put into making me a ' scientist'. I would also like to thank Dr. Fiona Schmiegelow for giving me tremendous support and motivation both in the field and at the office, Dr. Daryll Hebert and Alberta Pacific Forest Industries for their financial support, Dr. Tony Kozak for giving me statistical advice on numerous occasions, and Drs. Alton Harestad and John Richardson for serving on my graduate committee and providing much needed, valuable advice. Special thanks to Greg Cote, Carl Burgess, and Sarah Bunce for being excellent field assistants, and for keeping the energy positive and playful. I appreciate their strength in enduring all that mud! Also, thanks to Chris Brodie for putting up with me during the write-up and making dinner all those nights, fellow Centre for Applied Conservation Biology students for being so supportive over the years, and to Rana for making me get outside to walk, play, and laugh. vii CHAPTER I GENERAL INTRODUCTION Historically, ecological research about the management and preservation of amphibian populations has lagged well behind that of other taxonomic groups, especially mammals and birds (Gibbons, 1988). However, amphibian research has increased significantly in the past decade, in part from concerns over their declining populations worldwide, and in part because of recent mandates for preserving biodiversity. To be useful to resource management, much of the current ecological research on amphibians has focused on the proximate causes of species' declines, the effects of habitat disturbance, and amphibian habitat requirements. Amphibian populations fluctuate extensively from year to year (Berven, 1995), and this has sparked debate about whether declining amphibian populations witnessed today are as widespread or as serious as they appear (Pechman et. al., 1991; Blaustein et. al, 1994c). Because there are very few long term studies which capture the inherent variability of amphibian populations, our ability to determine the extent of this problem is hampered (Blaustein et. al., 1994c). Declines in amphibian populations have occurred worldwide in pristine, as well as disturbed, environments (Barinaga, 1990; Blaustein and Wake, 1990; Phillips, 1990). Evidence suggests that even species which were once considered numerous can decrease dramatically to critical population levels, or to extinction, within a relatively short period. Examples in western North America include the northern leopard frog (Ranapipiens) and spotted frog (Rana pretiosa), which have decreased rapidly in some areas within the last 20 years (Corn and Fogleman, 1984; Orchard, 1992). These trends concern scientists, because amphibians contribute significantly to the biomass of many ecosystems (Pough et. al., 1987) and may serve as good indicators of environmental health (Phillips, 1990). Due to the uncertainty surrounding 1 the stability of many amphibian populations at this time, guidelines established for their management should be conservative and cautious. Hypotheses advanced to explain amphibian population declines include increased ultraviolet radiation exposure at the larval stage (Blaustein et. al., 1994a), spread of a pathogenic fungal disease (Blaustein et. al, 1994b), overharvesting, predation from introduced species, and increased toxins in the environment (Blaustein and Wake, 1990; Phillips, 1990). One of the major causes of amphibian declines is believed to be the degradation and loss of habitat (Johnson, 1992). In North America, habitat loss for amphibians has largely resulted from the clearing and draining of land for agriculture, urban development, and from forest harvesting. In some areas, second-growth habitats which result from forest harvesting have lower species richness and abundance of amphibians than do late successional forests (e.g., Pough et al, 1987; Welsh, 1990; Petranka et. al, 1994; Dupuis et. al, 1995). The permeable skin of amphibians impose unique physiological constraints on their ability to exploit various habitats (Adolph, 1932). They are restricted to moist habitats which permit gas exchange and reproduction without dehydration (Zug, 1993). The forest environment, especially old-growth forests, provide protective cover and stable climatic conditions for amphibians (Heatwole, 1962; Welsh, 1990), but increasing land-use pressures on old-growth forests, as well as increasing proportions of second-growth habitat, may seriously threaten amphibian populations. The landscapes being created today have never existed before. The ability of organisms to adapt to increasing proportions of second-growth habitat, and fragmentation of old-growth forests, will influence community composition everywhere (Kellman, 1996). An example of an ecosystem facing vast changes in the coming decades is the boreal forest. This is one of the last large, relatively pristine ecosystems left in North America. 2 Nonetheless, there is increasing pressure to harvest boreal mixedwood forests for the pulp and paper industry. As this ecosystem is exploited, the landscape will become fragmented, and the landscape composition and configuration will be dramatically altered (e.g., increased proportions of early and mid-successional stands compared to old-growth forests, and cutblocks more rectangularly shaped, and of different sizes, compared to areas cleared naturally by fire). Concern over this dramatic large-scale harvest has prompted the initiation of a series of studies investigating the effects of timber harvesting on biodiversity in boreal mixedwood forests (Schmeigelow and Hannon, 1993; Stelfox, 1995). At Calling Lake, Alberta, a large collaborative study has been initiated to investigate the effects of fragmentation on biodiversity (Schmeigelow and Hannon, 1993). The majority of the work at Calling Lake focuses on birds, investigating species-area relations, breeding success, use of corridors, and habitat use. The research area contains four fragment sizes replicated three times each (i.e., 1, 10, 40 and 100-ha; surrounded by minimum 200-m wide clearcuts), as well as a large control area. Pre-treatment data were collected in 1993 when the research area was unharvested, and fragmentation occurred in winter 1993/94. In addition to the work being conducted on birds, effects of fragmentation on small mammal and amphibian populations have also been investigated. My research focused on wood frogs (Rana sylvatica), determining habitat associations of this species within boreal mixedwood forests, and effects of fragmentation. Amphibians are an ideal taxonomic group for investigating the effects of landscape level disturbances, such as forest fragmentation. Compared to many other vertebrates, amphibians have relatively small home ranges and low vagility, which means that they will have to adapt or perish in the face of large-scale habitat disturbance. Many amphibian species depend upon both terrestrial and aquatic environments, migrating between them on an annual basis, while others 3 may spend their entire life within one small area, such as a log (Zug, 1993). Both strategies are affected by forestry and habitat fragmentation. Those species which need to migrate between two or three habitat types depend upon unimpeded movement between these sites for their survival. Those species which move very little may not have the ability to successfully disperse to a new location following habitat disturbance, when, for example, microclimate conditions surrounding and within their log become unsuitable. The spatial distribution of suitable breeding sites, and of summer and winter ranges, will influence the distribution of amphibians within that environment. Forest practices change the configuration of suitable habitat types for amphibians (Welsh, 1990; Dupuis et. al., 1995). Whether amphibians are flexible in their use of habitats and movement corridors is unknown, but high flexibility is unlikely when many species are found to return to natal breeding sites and summer ranges (e.g., Bellis, 1965; Berven, 1995). As a result, landscape level planning of the juxtaposition of harvested and unharvested areas, especially in relation to wet patches, will likely be a key factor in management strategies which seek to maintain source populations of amphibians. Very little is known of the ecology of the amphibians which inhabit the boreal forest. Knowing what the potential impacts of forest harvesting may be for these species, and what their specific habitat requirements are, are essential for including amphibians in forest management guidelines. The objectives of this thesis were: 1) to determine the habitat associations wood frogs in the boreal mixedwood forest of Calling Lake, Alberta; and 2) to investigate effects of forest fragmentation on wood frogs within this ecosystem. From these findings, I make recommendations for further research that is needed to manage amphibian populations in boreal mixedwood forests. 4 CHAPTER H HABITAT ASSOCIATIONS OF WOOD FROGS (RANA SYLVATICA) EV BOREAL MIXEDWOOD FORESTS INTRODUCTION The boreal forest is a complex ecosystem comprised of a fine-scale mosaic of habitats (Peterson and Peterson, 1992). This complex mosaic of different serai stages and species composition is believed to have resulted from fire and insect disturbances, which occurred in irregular patterns and shapes across the landscape (Bonan and Shugart, 1989). Strong and Leggat (1981) described the boreal forest as having a variety of forest ecoregions, reflecting plant species' responses to changes in topography, soils, and climatic variables across the landscape. One of these ecoregions is referred to as the 'boreal mixedwood'. Alberta contains more than 50% of Canada's boreal mixedwood forest, which covers about 40% of the province and approximately 290 000 km2 (Rowe, 1972). In 1987 - 88, the government of Alberta leased 220 000 km to the pulp and paper industry without conducting an environmental impact assessment (Nikiforuk and Struzik, 1989). Species which inhabit this ecosystem have presumably evolved to cope with the specific conditions and disturbance patterns found there (Bunnell, 1995). Little is known of the habitat associations and requirements of the five species of amphibians which inhabit this ecosystem (western toad - Bufo boreas; Canadian toad - Bufo hemiophrys; northern leopard frog - Ranapipiens; wood frog - Rana sylvatica; and chorus frog -Pseudacris triseriata; Russell and Bauer, 1993). Amphibian habitat associations have been 5 studied in a wide range of environments throughout most of North America, but few studies have taken place in the boreal forest of Canada. In other parts of North America, numerous factors have been identified as influencing the presence, distribution, and abundance of amphibians. The main factors identified which influence the occurrence of terrestrial amphibians include the presence of standing water and/or moist air for breeding, rehydration, gas exchange, and escape from predators (Adolph, 1932; Pough et. al., 1987). In addition to standing water, the density, structure and type of terrestrial vegetation contribute to moisture levels on the ground and in the air (Geiger, 1965; Heatwole, 1962). Apart from the need for moisture, other habitat requirements identified for many amphibian species include abundant prey, hibernation sites, and the presence of appropriate structural attributes (e.g., coarse woody material) to provide camouflage and protection from extreme climatic conditions (Jaeger, 1980; Aubry et. al., 1988). Little is known of the habitat requirements of many amphibian species during the summer months away from breeding ponds, and especially during hibernation. The habitat attributes that amphibians require during these periods may be quite different from those they select for breeding, and may be just as critical to their survival. If breeding habitat alone is protected, only a portion of amphibian habitat requirements will be met. Management of forested ecosystems that support stable amphibian populations requires an understanding of all habitat needs. Three species of amphibian are found in the Calling Lake research area: the wood frog, the western toad, and the chorus frog. The wood frog was selected as an ideal study species for investigating amphibian habitat associations in the boreal mixedwood forest because of its dependence on both terrestrial and aquatic environments, relative abundance in the area, and ready trappability compared to the other two species. The objective of this study was to identify 6 habitat attributes of wood frogs in the boreal mixedwood forests of north-central Alberta, specifically 1) terrestrial attributes (i.e., vegetation, coarse woody material, canopy closure, and leaf litter depth), and 2) wetness attributes (i.e., percent wetness, number of wet patches, and variety of wet patches). Extensive timber extraction is occurring throughout the area and it is important that we identify key habitat attributes for various species, to plan appropriate juxtaposition of clearcuts, wet patches, and remaining forested habitat. This study provides information and recommendations for further research that will facilitate management decisions for maintaining stable amphibian populations in boreal mixedwood forests. RESEARCH AREA AND SAMPLING SITES The Calling Lake research area was established to study the effects of habitat fragmentation and timber extraction on biodiversity. The research area is located 250 km north 2 of Edmonton and encompasses approximately 120 km (12,000 ha) of boreal mixedwood forest. Early successional species in this ecoregion are trembling aspen (Populus tremuloides) and less importantly, balsam poplar (Populus balsamifera) and white birch (Betulapapyri/era). Late successional species are mainly white spruce (Picea glauca) and to a lesser extent balsam fir (Abies balsamea). Jack pine (Pinus banksiana) communities occur on sandy sites; black spruce (Picea mariana) dominate poorly drained areas. Wetlands comprise about one quarter of the region (Strong and Leggat, 1981). Historically, insect outbreaks and fire have played an important role in creating the mosaic of stand types found in the boreal forest (Bonan and Schugart, 1989). Al l of the study sites at Calling Lake are in old stands (80 to 130 years since 7 last disturbance) of aspen / mixedwood forest. The terrain in the research area has only minor variations in topography. The area has long, cold winters and relatively short, mild summers, with precipitation heaviest in July (Alberta Environmental Protection, 1996). Selection of the Calling Lake area for the fragmentation project was based on a number of criteria, including its representativeness of the boreal mixedwood ecoregion, the planned harvesting schedule (i.e., to occur in winter 1993-94), and its similarity to those areas where biodiversity research was being conducted by the Alberta Environmental Centre (Schmiegelow and Hannon, 1993; Stelfox, 1995). The research area was unharvested, and baseline data were collected in 1993. It was subsequently winter logged in 1993 / 1994, with replicates of various sized fragments created that were surrounded by minimum 200-m wide clearcuts. A large control area, > 3500-ha, was left unharvested. Fragment sizes of 1, 10, 40 and 100-ha were replicated three times. Time constraints and limited resources did not allow for sampling of amphibians in all study sites in 1993, thus only the 10-ha and 100-ha fragments (three replicates of each) and four control sites were sampled (n = 10). Preharvesting data were collected in July and August of 1993. In 1994 and 1995, six of the amphibian sampling sites were in isolated fragments, and sampling occurred from May to early August in both years. Four of the amphibian sampling sites were considered relatively 'younger' forest (80 - 90 years), and six were 'older' forest (110 - 130 years) according to Alberta Vegetation Inventory (AVI) mapping. Al l of the sites were primarily in aspen-dominated forest, however, some sites contained patches where the forest was dominated by white spruce (i.e., Sites C- l , T-2 , C-3; see Table 2.1 for site descriptions). 8 o I* "I 60 O > 1 co J u co es O O s I o co 5 3 & co 6 en 1' O of 1=1 CO l>> a> Pi T3 CD I o co CD ^ O 0 ) si •O ?\ g> 1 1 5 2 2 ~ ft 3 co s If SI CD 1 t-l So T 3 CD la * CO =tfc I I I "a B o " CD 0 ^3 1 ° 5 5 S o Is s* CD •s # B ° !3 -o •« '53 bo c3 o <;
P w C3 i rt i S •& C M P • 6 X 6 Pitfall Trap Grid <-25 m-> Four circular vegetation plots Figure 2.1. Layout of amphibian sampling sites, with trapping grid, four vegetation plots, and. wet patch search area. 11 Pitfall traps were made by removing the bottom from a large coffee can and attaching it to another intact can with duct tape. Holes for water drainage were punched into the bottom of the intact can. A plastic collar (1 lb. margarine tub with bottom removed) was inserted on the top to keep frogs from escaping, and a plastic lid on the collar allowed traps to be closed when not in use. Within 2 m of a flagged point, a hole was dug slightly larger than the diameter of the cans, using a post hole auger. The cans were placed snugly into the hole, flush with the ground. A small amount of moss and/or debris was placed on the bottom of the trap to provide shelter and to maintain moisture. A slit was cut into the plastic collar and a cord was hung into the trap to allow small mammals (e.g., shrews) to climb out and escape. Pitfall traps were left open continuously and checked every two to three days. In 1993, the traps were opened as the construction of each grid was completed, and the sampling period differed for each of the 10 sites in that year. I therefore used the average number of frogs caught per day for comparison of wood frog abundance across different sites. Information collected on each individual wood frog that I captured included the date, sampling site, trap number, mass (using a Pesola scale; visually estimated to within ±0.1 g), body length (using calipers, visually estimated to within ±0.5 mm; snout - urostyle with the frog pushed down slightly with the thumb to maintain consistency), and capture status (i.e., lack of toe clips indicated first capture, and a unique pattern of toe clips identified a recapture; see below). Size of individuals was recorded for two reasons: 1) to determine habitat associations of different size (age) classes ('new recruits' and 'adults'; see below); and 2) to eliminate 'new recruits' from capture data because they are quite transient and inclusion inflates abundance estimates in some areas (Heyer et. al., 1994). I used length data to distinguish the different size 12 (age) groups because I felt that it was measured more accurately in the field than mass (coefficient of variation around the sample mean was 23% - 28% for the length data from 1993 to 1995, and 73% - 85% for mass). Mass of individual frogs was likely affected more by duration of time spent in the trap than length due to exposure of individual frogs to different environmental conditions (e.g., length of time in the trap without food, bloating from sitting in flooded traps, etc.), and to measurement error (i.e., decreased precision of the 50 g Pesola scale at a low mass; many 'new recruits' were < 2 g). I created three data sets: the first consisted of all size classes combined (referred to henceforth as the 'combined' data set); the second included only 'adults' (i.e., > 25 mm); and the third included only 'new recruits' (i.e., < 25 mm) . The term 'adult' does not necessarily represent breeding individuals, only those that were not 'new recruits'. The 'new recruits' data set represented very small individuals, comprised largely of newly metamorphosed frogs from that year. The size of the 'new recruits' was easily determined by an increase in captures of individuals at or near 25 mm in length during late summer (Figure 2.2) and from the literature (Bellis, 1961; 1965). Bellis (1965) found that new wood frog metamorphs, one week before emergence from their aquatic environment, were approximately 20 mm in length. He also noticed a surge in capture rate of young of the year in late July and early August, with metamorphs first captured between July 18 and August 4 (Julian dates 199 to 216), depending on the year. Information on recapture rate and movements of wood frogs were obtained by marking and releasing trapped individuals (Ferner, 1979). I removed various toe tips using a marking system whereby each individual captured within a sampling site could be identified. The first individual caught within a site had the outer toe on the right hind foot clipped, the next 13 1993 50 • 45 E 40 • E 35 .c at 30 • c a> 25 • 20 • 15 < » 8 0 o A Y O U N G 04/19 05/09 05/29 06/18 07/08 date (month/day) 07/28 08/17 1994 15 04/19 05/09 05/29 06/18 07/08 07/28 08/17 date (month/day) 1995 15 • • • • • • • 04/19 05/09 05/29 06/18 07/08 07/28 08/17 date (month/day) Figure 2.2. Body length of wood frogs (snout to end of urostyle), caught between 1993 to 1995. Some individuals caught early in the season that were < 25 cm were likely young from the previous year. 14 individual had the second outer-most toe clipped, and so on. After the first 10 individuals had been trapped and clipped in this manner the next nine individuals had their right, hind outer toe clipped plus one other. The marking of the hind feet alone, with only two toes clipped, allowed for identification of 54 individuals. If > 54 individuals were caught in one site in a given year then three hind toes were clipped. This system was repeated in each sampling site in each year. Some species of amphibians regenerate clipped toes; however, Berven (1990) claimed that wood frogs do not regenerate clipped toes, and Bellis (1965) recaptured marked individuals across years. In 1993, frogs were not marked but toe clipping was used in both 1994 and 1995. Habitat and Vegetation Habitat attributes of each sampling site were measured in four, 0.04-ha vegetation plots located at the corners of each sampling grid, among the outer four pitfall traps (Figure 2.1). Vegetation was sampled using circular plots, transect lines, and quadrats, and was generally consistent with the methods used by other studies in the Calling Lake area (Martin, 1992). Two 22.6-m transects were placed at right angles in the centre of the vegetation plot. The ends of each transect defined the diameter of the largest circular plot. Within the largest circular plot (11.3-m radius), the number of trees of each species were counted and placed into one of four size classes (8 < 15, 15 < 23, 23 < 38, 38+ cm dbh). Within a smaller nested circular plot (5-m radius), the number of saplings and poles (< 8 cm dbh) of each tree species was counted. A 1-m quadrat was placed at the centre of the circular plot and at the 5-m mark along each of 4 transect lines. In the centre quadrat, the number of stems of each shrub species (with stems taller than 50 cm) were counted. Within the remaining four quadrats percent ground cover was estimated visually to the nearest 5% for seven categories; all green vegetation, 15 grasses/sedges, shrubs, forbs, leaf litter, woody material (all sizes), and mosses/ferns. Litter depth (cm) was measured at the 3-m mark along each transect line (i.e., four times per vegetation plot). Coarse woody material (CWM; all sizes) was measured along each transect line, and placed into four size classes (CWM1 = 0<2, CWM2 = 2<5, CWM3 = 5<11, CWM4 = 11+ cm diameter). Any piece of downed wood which crossed the transect line was included, and the diameter measured at the point of intersection. Canopy closure was measured by ocular estimation within the four outer pitfall traps (i.e., overlaying the vegetation plot). A 40-cm long piece of PVC piping (2-cm diameter) was used to sight canopy closure directly above the researcher's head. A total of 30 points were distributed evenly between the four pitfall traps and recorded as either open or closed canopy, with open indicating that no foliage was in view, and partial or complete foliage cover classed as closed. The researcher walked systematically between the four traps following five transect lines (running north / south) placed 5 m apart, and measures were taken every 4 m along each line (six measures per line). A measure of canopy closure was then calculated by taking the ratio of closed sites to total sites measured and recorded as a percentage (e.g., 20 closed sites out of 30 sites equals 67% canopy closure). Wet Patches In 1994, wet patches were first located by walking systematically through the entire sampling site (325 by 325 m; Figure 2.1). Two researchers walked 25 m apart and noted the location of every 'wet' patch, which ranged from saturated soils (water appeared under foot pressure) to standing water, and the approximate size (area in m ) of each patch was measured. The broad definition of wet patches was designed to include those which are wet only during 16 part of the season. Wet patches retain water for different periods during spring and summer depending upon microsite (e.g., seep, depression, permanent, temporary, etc.). Wet patches were first located and measured in late June to late July in 1994, which coincides with the wet season in north-central Alberta (Alberta Environmental Protection, 1996), and were re-measured at three times in 1995, in late April / early May, June, and July. The three repeated measures in 1995 provided a more accurate estimate of average wetness of each sampling site than did the single measure in 1994. The approximate size (area in m2) of each wet patch was measured in the field and from this the percent wetness of the sampling site was derived (i.e., total area of all wet patches within each sampling site divided by the total search area, multiplied by 100). From these data, two independent variables were derived based on the mean from the three 1995 measures; percent wetness of the sampling site, and number of wet patches in the sampling site. Different types of wet patches were observed in the field from broad microsite characteristics. To determine whether or not wood frogs were associating with different types of wet patches, six wet patch classifications were created based on two broad criteria. The first criteria, main vegetation component of the wet patch, had three classes; 1) the patch contained mostly trees throughout, with extensive canopy closure; 2) the patch contained mainly shrubs throughout, with few or no trees; and 3) the patch consisted of mainly grasses throughout, with no shrubs or trees. The second criteria, duration and extent of saturation, had two classes; 1) the patch was mainly dry all, or most, of the time; and 2) the patch was mainly wet all, or most, of the time. 17 Data Analyses I conducted analyses on only those sites, and for periods, where fragmentation was not a confounding factor influencing amphibian abundance (i.e., all 10 sites in 1993, and the four control sites over the three-year sampling period). The four vegetation plots within each sampling site were analyzed separately to capture the greatest variability and detail in habitat attributes. These analyses were largely exploratory in nature, because the objectives were not proposed before the sampling design was established (improvements to the sampling design are mentioned in the discussion and in Chapter IV, Conclusions and Recommendations). To investigate habitat associations of wood frogs in relation to terrestrial attributes (i.e., Objective 1), I used forward stepwise multiple regressions (SMR) and multivariate discriminant function analyses (MDA). I conducted a SMR on capture rates as a data reduction technique for control sites over the three sampling years. Using this technique for reduction ensures inclusion of all variables that are potentially biologically important in influencing abundance of wood frog (Myers, 1986). Assumptions for regression were likely not met due to the presence of numerous zeros in the data, therefore the SMR was used as a data reduction technique only and significance values for the model itself were ignored. I used data gathered from the 16 vegetation plots (4 control sites with 4 vegetation plots each) and the four pitfall traps immediately surrounding each plot. Only data from these four pitfall traps were used because they should reflect abundance of wood frogs within the vegetation plot more closely than data from the rest of the trapping grid. The total number of wood frogs caught over the three years within each group of four traps was the dependent variable. This was regressed against 51 independent habitat variables (see Appendix 1). If a variable was found in less than 25% of the 18 vegetation plots it was excluded from the analyses for all plots. The stepping procedure was terminated when an insignificant variable entered the model (i.e., a > 0.05). Once independent variables were chosen by SMR, I conducted a series of MDAs to determine how well I could predict the presence or absence of wood frogs in 1993 based on these variables. I used only the pre-treatment 1993 data for the MDA to avoid any confounding effects of fragmentation, and to include a broader range of sites than those provided by only the control sites for 1994 and 1995. Capture rates in 1993 were relatively low when only four traps for each plot were used in the analysis, and the plots were easily assigned a present or absent value. The total number of captures per plot fell into four categories, with 28 plots having zero captures, and the remaining 12 plots containing one (n=9), two (n=l), or four (n=2) wood frogs. Al l plots in which frogs were captured were placed into the presence category, regardless of how many were caught, because the sampling period for each plot was uneven that year and the capture rates were low. Al l of the chosen SMR variables were entered into a MDA together to determine how well the variables discriminated as a group. Each of the variables was then entered into a stepwise MDA (S-MDA) one at a time, in the order in which they were chosen by the SMR, to determine which subset of variables discriminated the best. Results from an MDA are always optimistic when the data used to derive discriminant functions are used to test the same data set, therefore, an N-way validation was done on half of the plots with the most significant discriminating variable (Tabachnick and Fidell, 1983). The four vegetation plots per sampling grid were used as groupings of the complete data set, and two groups of 10 plots were randomly selected for validation (i.e., vegetation plots 1 and 4). Each of the two groups was left out of the MDA, one at a time, and the remaining 30 plots were used to determine new discriminant 19 functions, which the group of ten was then tested against. Once the variable list had been reduced and the most significant discriminating variable found from the MDA, a Pearson correlation analysis was conducted to determine which independent variables were correlated most highly with this discriminating variable. I repeated the SMR and MDA for the 'adult' data, then compared results from the two sets of analyses. Pearson correlation analyses were conducted to investigate the relationship between the abundance of wood frogs and wetness of the sampling sites (i.e., Objective 2). The number of 'adult' and 'new recruits' caught in control sites in 1995 were used as dependent variables, and the average number of wet patches and the average percent wetness of control sites, based on the three sampling periods in 1995, were used as independent variables. I also examined the relationship between the density (i.e., number of each type) and the variety (i.e., number of different types) of wet patches to the abundance of wood frogs within control sites. A significance value of 0.10 was used for all statistical tests to reduce the probability of committing a Type II error, and all analyses were conducted with SYSTAT, Version 5 (Systat Inc., 1992). RESULTS Capture Rates The capture and recapture rates of wood frogs in the Calling Lake area were quite low each year, especially in 1993, when sampling occurred for the shortest period (2-3 weeks; Table 2.2 shows capture rates of control sites). The rankings of the sites, in terms of high and 20 Table 2.2. Wood frog capture rates in control sites from 1993 to 1995 at Calling Lake, Alberta. 1993 1994 1995 Site Mean No. . Mean No. Recaptures Mean No. Recaptures Total Mean No. Caught Caught Caught Caught Caught Per Day Per Day Per Day Per Day C - 1 0.21 0.18 0 0.23 0 34 0.21 Control (0.00)* (0.15) (0.16) (0.10) C - 2 0.07 0.10 0 0.25 0 26 0.14 Control (0.07) (0.06) (0.12) (0.08) C - 3 0.06 0.11 0 0.38 1 (15 days apart) 36 0.18 Control (0.06) (0.05) (0.38) (0.16) C - 4 0.06 0.07 0 0.38 2 (from 1994) 33 0.17 Control (0.00) (0.06) (0.35) 1 (21 days apart) (0.14) Mean 0.10 0.12 0.31 0.18 (0.03) (0.08) (0.25) (0.12) data within brackets indicate 'adult' data set. 21 low capture rates, fluctuated from year to year. Recapture rates were too low for detecting trends across the sites. Among the four control sites for all three years combined, two had relatively higher capture rates of 'adults' compared to 'new recruits'; the other two sites had higher captures of 'new recruits' than 'adults'. Habitat and Wet Patch Descriptions for Sampling Sites The ten sampling sites were located within two different age classes of forest, as determined by Alberta Vegetation Inventory mapping (AVI). Sites T- l , H - l , C - l , and C-2 were in the younger age class (approximately 80 - 90 years old), and all other sites were within the older age class (approximately 110 - 130 years old). Vegetation plots within older control sites (i.e., C-3 and C-4) generally had higher tree and shrub species richness than did younger sites (Figure 2.3). Sites within these age classes also tended to differ in the density of trembling aspen (Figure 2.4); the younger sites had greater numbers of stems, especially in the first two size classes of trembling aspen, and the older sites generally had lower densities. The older sites also became more diverse in age class of trees, and they tended to have a greater mixedwood component, with increasing numbers of white spruce and paper birch entering into the subcanopy and canopy (Lee et. al., 1995). The older sites had higher shrub densities, with shrubs often reaching > 1.5 m in height, as well as the presence of numerous gaps in the tree canopy. Furthermore, the younger sites tended to be relatively dry compared to the older sites, which were split between being either very wet (Sites T-2, H-2, C-3, and C-4), or relatively dry (Sites T-3 and H-3; Figure 2.5). Percent wetness ranged from 0 - 12% across the sites in 1994, and from 0 - 20% among the three sampling periods in 1995. A comparison of the 1994 results to the last two samples in 1995 (overlapping periods for both years) indicated that all but two sites (Sites T-2 and C-3) 22 a) Tree species richness 5.0 r