Northern bog lemmings: survey^ population parameters^ and population analysis A Report to: USDA Forest Service Kootenai National Forest 506 U.S. Highway 2 West Libby,MT 59923 Submitted by: James D. Reichel and Janelle G. Corn April 1997 Montana Natural Heritage Program 1515 E. Sixth Avenue P.O. Box 201800 Helena, MT 59620-1800 © 1997 MontanaNatural Heritage Program This document should be cited as follows: Reichel, J. D. and J. G. Com. 1 997. Northern bog lemmings: survey, population parameters, and population analysis. Unpublished report to the Kootenai National Forest. MontanaNatural Heritage Program. Helena, MT. 27 pp. ABSTRACT Northern bog lemmings {Synaptomys borealis) were discovered in 1 992-3 in single patches within the Cody Creek and South Fork Hawkins Creek drainages. During the 1 994 field season we surveyed these two drainages to determine the number, size and location of other suitable habitat patches. No suitable habitat patches larger than about 50 m^ were located. The total number of known bog lemming sites in Montana is 1 8, the most sites in any of the lower 48 states. Known sites in Montana range in size from 1 to approximately 340 acres. The best habitat predictor for potential northern bog lemmings sites in Montana is the presence of large, thick moss mats, particularly sphagnum moss. ACKNOWLEDGMENTS We would like to thank Bob Summerfield and Seth Diamond for their help throughout the study. D. E. Pearson provided access to his invaluable unpublished information. S. W. Chadde, S. V. Cooper, J. C. Elliott, and B. L. Heidel identified plants and plant communities. Help with field work and other logistical support was provided by G. Altman, J. Berry, A. Bratkovich, and other Forest Service personnel. M. Beer, D. Dover, C. Jones, and K. Jurist helped with database applications and mapping in the preparation of the report. L. S. Mills and J. Caratti provided helpful criticism of earlier versions of a P VA model. Financial support for the proj ect came from the Kootenai National Forest (U. S . Forest Service) and the Montana Natural Heritage Program (Montana State Library). TABLE OF CONTENTS ABSTRACT iii ACKNOWLEDGMENTS iii INTRODUCTION 1 METHODS AND MATERIALS 3 RESULTS 6 DISCUSSION 12 STATEWTOE MANAGEMENT RECOMMENDATIONS AND RESEARCH NEEDS 18 LITERATURE CITED 20 APPENDIX I. NORTHERN BOG LEMMING PVA MODEL DESIGN PARAMETERS 23 APPENDIX IL MUSEUMS CONTACTED AND RESULTS 27 LIST OF TABLES AND FIGURES Table 1 . Timing of reproduction in northern bog lemmings, as estimated from museum specimens 8 Table 2. Table 2. PVA results for the northern bog lemming on Sunday Creek 11 Table 3. Plant communities present at 6 northern bog lemming sites 13 Table 4. Characteristics of known bog lemming sites, plus several additional sites in the Sunday Creek complex, in Montana 14 Figure 1 . Map of Northern Bog Lemming locations in Montana 2 Figure 2. Sunday Creek bog lemming complex showing potential and known habitat patches 5 Figure 3 . Distributions of time to quasi-extinction (<1 00 individuals) over 50 years (1 00 simulations) in the Sunday Creek metapopulation of northem bog lemmings 10 INTRODUCTION The northern bog lemming (Synaptomys borealis), a small, grayish brown, vole-like microtine, is related to the true arctic lemmings (Lemmus). Nine poorly differentiated subspecies are currently recognized (Hall 1981). The northern bog lemming has a total length of 1 1 8- 1 40 mm including its very shorttail (19-27 mm) (Banfield 1974,Hall 1981). The combination ofa tail less than 28 mm long and a longitudinal groove in the upper incisors distinguish the northern bog lemming from all other mice found in Montana. The northem bog lemming is boreal in distribution, occurring in North America from near treeline in the north, south to Washington, Idaho, Montana, Minnesota, and New England. It typically inhabits sphagnum bogs and fens, but is also occasionally found in other habitats including mossy forests, wet sub-alpine meadows, and alpine tundra. One subspecies (S. b. artemisiae) lives on sagebrush hillsides in eastem British Columbia (Anderson 1 932). The northem bog lemming is rarely trapped and is one of the least known mice in North America. It is listed as a Species of Special Concern by the Idaho, Minnesota, Montana, and New Hampshire Natural Heritage Programs, and on the Special Animal Priority List of the Washington Natural Heritage Program. The subspecies Synaptomys borealis artemisiae is listed as a Species of Special Concern by the British Columbia Conservation Data Centre. A few relict populations occur in the lower 48 states; the subspecies chapmani occurs in Montana, Idaho, and northeast Washington (Hall 1981). Bog lemmings are known from 4 locations in Idaho and 8 in Washington, all from within 80 km of the Canadian border (Johnson and Cheney 1 953, Wilson et al. 1980, Reichel 1984, Groves and Yensen 1989, D. Johnson pers. comm.). Prior to 1992, evidence of bog lemmings in Montana included: 1)6 locations on the west side of Glacier National Park (Wright 1 950, Weckwerth and Hawley 1 962, Hoffmann et al. 1 969, Pearson 1 99 1 ); 2) Shoofly Meadows in the Rattlesnake drainage north of Missoula (Adelman 1 979), and 3) a single skull recovered from a Boreal Owl (Aegoliusfunereus) pellet west of Wisdom (J. Jones pers. comm.); where the owl captured the lemming was unknown. In 1 992 and 1 993, 5 1 sites were trapped which located 1 new populations of northem bog lemmings (Figure 1 ) (Reichel and Beckstrom 1 993 , 1 994). The Maybee Meadows site is the southem-most known population of the species outside of New England and one of two Montana populations known from east of the Continental Divide. All 1 sites found in 1 992- 1 993 were associated with thick mats of moss. Their disjunct distribution and rarity may be due to : 1 ) the localized nature of their primary habitat; and 2) their currently patchy distribution from more widely distributed populations during the Pleistocene (a glacial relict). Species like the northem bog lemming — ^rare, patchily distributed, confined to rare habitats — are at particular risk of extinction (Shaffer 1981). Population viability analysis (PVA) is one means of assessing apopulations=3 risk of extinction quantitatively. PVA models that incorporate demographic and environmental stochasticity (uncertainty) in population trend are particularly powerful analytical tools, and are available in user-friendly computer programs. Stochastic PVAs require extensive background information on the population=3s demographic characteristics (fecundity and survival rates). Unfortunately, little is known about northem bog lemming life history and demography. A few notes in the literature indicate litter sizes vary from 3-8, with 2 (or possibly more) litters Figure 1 . Northern Bog Lemming occurrences in Montana. Locations are from Wright (1 950), Weckwerth and Hawley (1962), Adelman (1979), Pearson (1991), Reichel and Beckstrom (1993), and this report. ^ Sites discovered in 1993 I I Known sites pre 1993, typical habitat (^ Known sites pre 1993 , atypical habitat ^^L National Forest 10 20 Miles per year. It has been suggested that some individuals breed the same year they are bom (perhaps 60- 90 days old). Most reproduction information is scattered throughout a literature that deals mainly with distribution. More is known about the northern bog lemming=5s congener, the southern bog lemming (S. cooperi). Southern bog lemmings are distributed in eastern and central North America, from southeastern Canada west to Minnesota, and south as far as Kansas and North Carolina. They inhabit a wide variety of habitats, but, like northern bog lemmings, tend to be associated with sphagnum bogs in eastern forests (Linzey 1 983). Population densities also vary widely, from 4 - 5 1/ha (Linzey 1 983). Detailed population studies conducted for southern bog lemmings in Kansas (Gaines et al. 1 977, 1 979), Illinois (Beasley and Getz 1 986), and Virginia (Linzey 1 983) indicate that breeding occurs all year in the southern part of its range, but ceases in winter in the northern part of its range. We used these data to estimate population densities and demographic parameters in a P VA for the northern bog lemming. The lack of species-specific data makes the model necessarily preliminary, but nonetheless it can be used to suggest what additional biological information is essential to develop a sound bog lemming management plan. Amulti-year study of northern bog lemmings in Montana was begun in 1 992. Objectives during 1994 included: 1) Determine the extent of the suitable habitat in the Cody Creek and Hawkins Creek drainages. 2) Review the available information on northern bog lemmings and closely related voles to develop a model of population viability; we will discuss which parameters are weakest and what additional data are necessary to strengthen our confidence in the model. 3) Determine what additional biological information is most critical for development of a bog lemming management plan. This report also contains a summary of habitat characteristics at bog lemming capture locations in Montana. METHODS AND MATERIALS Surveys We visited Cody Creek and Hawkins Creek drainages in western Montana, examining riparian habitats to determine their suitability for northern bog lemmings. We walked much of both drainages within three miles of the known bog lemming areas, and used aerial photos and USGS 7 V2 -minute topographic maps to find locations of potential bog lemming habitat. All areas within the drainages which looked potentially suitable were examined. PVA Model Development We evaluated documentation for several of the PVA computer programs most widely available for use on a PC, and chose RAMAS/GIS (Applied Biomathematics, Setauket, NY) for the analysis. This program uses an age- or stage-based population growth matrix, includes stochastic variation in population change, and accommodates metapopulation dynamics, with movement of individuals between habitat patches, or populations (Akcakaya 1993). Appendix I defines parameters in the population growth matrix. To estimate values for the matrix parameters, we used information from northern bog lemming museum records, and data from demographic studies of southern bog lemmings. We obtained museum data by contacting all museums with medium to large mammal collections in the U.S. and Canada. They were asked to provide data on all northern bog lemmings in their collections including catalog number, date, sex, weight, location, and collector. Additionally they were asked to provide any data on specimens which had reproductive information associated with them; for females: embryos, placental scars, perforate/imperforate vagina, lactation; for males: testis or seminal vesicle size, scrotal/inguinal testes. We estimated the timing and duration of the breeding season, and the proportion of females breeding by summarizing museum data to determine the total number of females collected per month, and the number and percentage of pregnant females each month. For each breeding interval, we estimated litter size from embryo counts, with separate estimates for subadults and adults. Age class was determined from museum records, or was based on body mass, a reliable indicator of age in the southern bog lemming (Gaines et al. 1 997). We estimated survival rates of subadults and adults from demographic studies of southern bog lemmings (Beasley and Getz 1 986, Gaines et al. 1 977). We used these parameter estimates in a stage-structured population projection with 3 stages (subadults, adults breeding for the first time, and adults breeding for the second time), with a pre-birth pulse and 2 litters per year (Appendix I). The projection interval (i.e. the time step for calculating survival) is 3 months, the average interbirth interval. We chose the complex of 25 habitat patches on Sunday Creek, of known or potential use by northern bog lemmings, as the basis for the PVA (Figure 2), with each >5patch=3 defined as a population, and the entire complex in the Sunday Creek drainage defined as the metapopulation. For the preliminary metapopulation analysis reported here, we assumed that all patches were of equal habitat quality (i.e. population density did not vary between patches), with population sizes varying only according to patch size. Thus, even though the small bog habitats (>5good-excellent habitat=5 patches in Figure 2) are thought to be the primary habitat for northern bog lemmings, their importance relative to lower quality, but typically larger, habitat patches is not quantified in the PVA, because data for that type of simulation are not available. The implications of the assumption of equal habitat quality will be discussed in later sections of this report. The initial population size of each population of bog lemmings in Sunday Creek was determined by the areal extent of each habitat patch (from MNHP records) and density estimates for southern bog lemmings. We estimated initial population density as 12 lemmings/ha, the density of southem bog lemmings at the margin of their range in forested habitats of northeastern US, comparable conditions to northern bog lemmings in MT. The initial population stage structure was assumed to be a stable age distribution, and was calculated by the program. We simulated population growth as an exponential growth function to a ceiling (K, carrying capacity; Appendix I). K was calculated from the maximum densities reported for southem bog lemmings (5 1/ha; Gaines et al. 1 977, 1 979). Stochastic variation in population growth rate was estimated using the standard errors of estimates of average litter size, proportion of females breeding, and survival obtained from museum records or the literature. We assumed that environmental stochasticity is completely correlated across populations, because the Sunday Creek drainage is an area of limited extent likely experiencing similar environmental conditions across populations. Figure 2. Sunday Creek bog lemming complex showing potential and known habitat patches. Good - Excellent Habitat I I Fair - Poor Habitat Unkown Habitat ^ Bog Lemming Capture Site Movements of animals between populations were assumed to be constant across age classes at 0. 1 0. Dispersal rates were distance-dependent, as defined by a distance fiinction in the RAMAS/GIS program (Akcakaya, 1 993). Distances between populations were calculated fi-om the center of each population by the program, using UTM coordinates of each population. The program calculates dispersal of survivors, so each estimate of survival rate was increased by the dispersal rate to allow for dispersal loss from populations. We assumed all age classes were equally likely to disperse, and initially set the average dispersal distance at 1 00 m and the maximum dispersal distance at 500 m. PVAModel Output Simulations were initially run for short intervals (10 years with 50 replicates) to determine whether population parameters were reasonable; the long-term projection was 50 years with 1 000 replicates per simulation. RAMAS/GIS reports a variety of simulation results (Akcakaya 1 993); we report (1) final metapopulation occupancy (average [S.D.] number of extant populations [occupied patches] ; (2) quasi-extinction risk (probability [95 % CI] that the metapopulation will fall below 1 00 individuals at the end of the simulation) (3) time to quasi-extinction (median time of quasi-extinction, distribution of extinction times, and cumulative extinction probability). PVAModel Modifications The basic simulation described above was modified to test the effects of dispersal distance and life-history variables on the outcome of the PVA. Average dispersal distance was increased from 1 00 m to 500 m, and maximum dispersal distance from 500 m to 5 km. Additionally, the original values for adult fecundity or survival probability were increased by 0. 1 and subsequently decreased by 0. 1 and PVA outcomes compared to test the effects of their variation on PVA outcomes life-history variables most sensitive to change. Differences between simulations in quasi-extinction risk were tested using the Kolmogorov-Smimovtest statistic (Akcakaya 1993). RESULTS Surveys One of us (J. Reichel) visited the Hawkins Creek drainage on 1 1 September 1 994, and the Cody Lakes area 8 September 1 994. Both areas appeared to be of low potential for northern bog lemming habitat. No suitable habitat patches larger than about 50 m^ were located. This extent is substantially smaller than the smallest habitat patches used by bog lemmings in Montana (1 ac. or approximately 4,046 m^). The sites examined were not trapped in 1 994, however. PVAModel We contacted 39 museums for information pertaining to northem bog lemmings in their collections (Appendix II). Twenty-three museums sent us data on a total of 484 animals from 1 6 provinces and states. Specimens from Alaska (139) and Manitoba (94) dominated the collection; specimens from Alberta (42), British Columbia (37), Quebec (64), and the Yukon (47) were also well represented in collections. The remaining 1 states and provinces were the source of fewer than 20 lemmings each, 8 of those states or provinces were the source of fewer than 10 specimens. The distribution of reproductive activity from museum specimens suggests a 6 month breeding season with 2 litters per year, April - June, and July - September (Table 1 ). The earliest date any museum specimen was captured which had embryos was 27 April and the latest date was 1 October. However, one female weighed only 1 1 .8 g on 3 1 March, which indicates that at least some breeding takes place over the winter. In the museum specimen sample, breeding prevalence was lower and litter sizes were smaller on average for subadults than adults, and litter sizes for adults was smaller for the second litter than the first litter of the year. For subadults, the proportion of females breeding in the first litter of the year was 0.57 (4 of 7), and litter size averaged 3.5 (SE = 0.5, N = 4); in the second litter of the year, the proportion of females breeding was 0.125 (5 of 40), and litter size averaged 3.25 (SE = 0.829, N = 5). For adults, the proportion of females breeding in the first litter of the year was 0.75 (18 of 24), and litter size averaged 4.28 (SE = 1 .63, N = 1 8); in the second litter, proportion of females breeding was 0.4848 (32 of 66), and litter size averaged 3 .9 (SE = 1 . 1 , N = 32). Beasley and Getz (1 986) found that individual survival probability from birth to adult age in the southern bog lemming was 0.068 1 (SE = 0.0575). Correcting the survival rate for dispersal (increasing the estimated survival rate by 0. 1 ) increased the survival rate to adult age to 0. 1 68 1 . We assumed that survival from birth to adult is constant over the interval birth to subadult and subadult to adult, and estimated the rates of survival of each of these two intervals as (0.1681)^^^, or 0.4100. Southern bog lemmings have an average adult 2-week survival rate of 0.7067 (SE = 0.499) in summer, and 0.7683 (SE = 0.060 1 ) in winter (Gaines et al. 1 977). Survival over the prediction interval (3 months), is the 2-week survival rate multiplied 6 times, or 0.7067^ (=0. 1245) in summer, and 0.7683^ (=0.2057) in winter. Dispersing adults were taken into account by adding 0. 1 to summer and winter adult survival rates, increasing them to 0.2245 and 0.3057, respectively. These estimates of life history parameters were used to calculate variables used in the in the Leslie matrix (see Appendix I for details). The resulting Leslie matrix is: 0.492 1.3161 0.7752 0.4100 0.3166 Demographic and environmental stochasticity was modeled from the standard errors of the estimates for survival, litter size, and proportion of females breeding used in the population matrix (see Appendix II). The Leslie matrix for stochasticity is: 0.1659 0.3910 0.2639 0.2399 0.0419 Table 1 . Timing of reproduction in northern bog lemmings, as estimated from museum specimens. Month January number of female specimens number with embryos noted percent with embryos February March April 2 1 50 May 12 9 75 June 25 12 48 July 38 15 39 August 72 17 24 September 26 3 11 October 3 1 33 November December Total 178 58 33 PVAModel Output The stable age distribution for northern bog lemmings, calculated by RAMAS/GIS from the Leslie matrix, was dominated by subadults (66 %), compared to 34 %older adults breeding for the first or second time. A preponderance of subadults in the trappable population is observed in small mammals during the breeding season (pers. obs.), suggesting that the Leslie matrix is a reasonable estimate for this species. The finite rate of population increase (X [lambda]) predicted by this Leslie matrix is 1 .0755, which indicates an increasing population (A>1). However, when stochasticity is introduced into the population projection, population growth may not occur. From the basic model, with average and maximum dispersal distances of 1 00 m and 500 m, respectively, only 5.4 populations (N=1000 replications) remain occupied after 50 years. The probability of quasi-extinction (<1 00 individuals) in the metapopulation is 0.257 at 50 years, and the median time to quasi-extinction is 26.2 years (Figure 3, Table 2). PVAmodel modifications Table 2 summarizes the outcome of the population projection for the Sunday Creek metapopulation under the various modeling scenarios. By increasing average distance moved from 1 00 m to 500 m, and maximum distance from 500 m to 5 km, probability of quasi-extinction increased and median time to quasi-extinction decreased over 50 years relative to the simulation with shorter dispersal distances (Table 2). However, the average number of occupied patches after 50 years was higher with increased dispersal distance (9.2). The results of manipulation of fecundity and survival rates indicates that the model is sensitive to both variation in fecundity and survival rates, but variation in fecundity rates may affect population projection more than to those of survival rates (Table 2). When adult fecundity rates were decreased by 0. 1 0, the probability of quasi-extinction more than doubled, and median time to quasi-extinction decreased by almost 1/2 half (Table 2); when fecundity was increased, quasi-extinction probability decreased accordingly. In contrast, increasing or decreasing adult survival rates by 0. 1 had less pronounced, but still highly significant, effects on quasi-extinction risk. All simulations resulted in relatively high quasi-extinction probability (Akcakaya 1 993). Figure 3 . Distribution of probability to quasi-extinction (<1 00 individuals) over 50 years (1 00 simula- tions) in the Sunday Creek metapopulation of northern bog lemmings. Vertical bars indicate probability of quasi-extinction during a given year; continuous solid line is cumulative quasi-extinction probability; continuous dashed lines are 95% confidence intervals. 0.8 c _o o X LU I O H — o 05 O Median time to quasi-extinction 20 26.2 30 Number of Years Table 2. PVAresults for the northern bog lemming on Sunday Creek. The basic simulation modeled the population projection with average and maximum dispersal distances of 1 00 m and 500 m, respectively. The details of the remaining modifications to the model are described in the text. Difference test results are Kolmogorov-Smimov test statistic D and significance level ("^=0.05, "^ "^=0.0 1 , * * *=0.00 1 ) for difference between the basic model and each model modification. Simulation Average number (s.d.) of extant populations after 50 years Median time (years) to quasi- extinction Extinction probability after 50 years Difference test D signif. Basic 5.4 ( 6.3) 26.2 0.257 - - Increase dispersal distance 9.2 (9.8) 22.1 0.331 0.08 ** Increase fecundity 10% 8.6 (7 ) 36.2 0.158 0.19 *** Decrease fecundity 10% 2.6 (4.6 ) 16.6 0.572 0.33 *** Increase survival rate 10% 7.1 (7.2) 28.2 0.258 0.1 *** Decrease survival rate 10% 3.7 19.5 0.459 0.2 *** 11 DISCUSSION Distribution . While northern bog lemmings were not found in the surveyed areas, it is within their range in Montana which includes the northwest comer of the state east to the Rocky Mountain Front, south through the mountains to Lost Trail Pass on the Continental Divide (Figure 1 ). The Maybee Meadows site is the southern-most site known for the species outside of New England; two sites in New Hampshire are about 1 60 km farther south (Clough and Albright 1 987; Reichel and Beckstrom 1 993, 1 994). The Maybee Meadows and Wood Creek sites are the only known northern bog lemming sites east of the Continental Divide in Montana. We expect additional populations will be found across western Montana, perhaps as far south as Yellowstone National Park, and possibly east to mountain ranges such as the Little Belt Mountains. The known elevation range for Montana is from 1018 m (3340 ft.) (McDonald Creek, Pearson 1991) up to 1987 m (6520 ft.) (Maybee Meadows, Reichel and Beckstrom 1993). Detectability . During 1 992- 1 993 lemmings were found at 1 of 1 7 sites that appeared to have suitable lemming habitat. Either lemmings were at 7 of those sites and we failed to detect them, or we sampled 7 sites with apparently good habitat which lacked lemmings. A combination of the two is a possibility (Reichel and Beckstrom 1 993 , 1 994). The percentage of sites with good habitat which had lemming captures was slightly higher than that of Pearson (1991) who found lemmings at 3 of 1 1 bog/fen sites trapped with Sherman live traps in 1 989-90. Habitat Patches . Bog lemmings have been found in at least nine community types (Table 3). However, peatland communities constitute a very small proportion of the landscape in Montana and have not been adequately classified (Bursik and Moseley 1 992). Whether new information on these fens will result in newly defined community types which closely approximate habitat used by northem bog lemmings remains to be seen. Extensive thick moss mats were present in all but one of the lemming sites found during our previous surveys (Reichel and Beckstrom 1 993, 1 994), and were also present at Numa Ridge Bog, McGee Meadows (Pearson 1991, P. Lesicapers. comm.) and Shoofly Meadows (Pearson 1991, S. Chaddepers. comm.). In 1 993 J. Reichel spent several hours along Camas Creek in the vicinity of the first lemming population known from the state (Wright 1 950) and found only scattered clumps of moss. Weckwerth and Hawley (1 962) did not adequately describe the two specific sites where they captured bog lemmings, but they were visited by D. E. Pearson (pers. comm.) who found that they were not located in fens or covered by thick moss mats. At these three sites trapping was conducted multiple years, often twice each year (Camas Creek: 1 8 years [Hoffmann et al. 1 969] ; Anaconda #1 : 6 years spring and fall [Jonkel 1 959] ; Anaconda #6: 4 years spring and fall [Jonkel 1 959]). Despite this intensive trapping, only 3 individuals have been taken in Camas Creek in 2 of 1 8 years, and 1 individual at each of the two Anaconda Creek sites. A similar situation exists with the McDonald 12 Table 3. Plant communities present at 6 northern bog lemming sites. CommunitvWDhase Sunday Creek Cody Lakes Bowen Creek Wood Creek Maybee Meadows Meadow Creek Abies lasiocarpia WCalamagrotis canadensis yes Picea WSalixgeyeriana- Carex utriculata yes Salix drummondiana yes Salix planifolia- Salix wolfii WCarexaquatilis yes Betulaglandulosa WCarex utriculata yes Betulaglandulosa- Eleocharis pauciflora WCarex lasiocarpa yes Betulaglandulosa- Carex lasiocarpa yes Carex utriculata yes yes (=C. rostrata) Eleocharis pauciflora yes 13 Table 4. Characteristics of l5catastrophic event=5, the removal of the central population >3Sunday 4' (Figure 2). Removing Sunday 4 from the metapopulation analysis did not increase the likelihood of extinction, as we would expect based on the quality of the habitat.. Rather, probability of extinction declined by 0. 1 0. Population density of Sunday 4 (actually 2 small patches) is low, because population size in the simulation was based only on areal extent of patches. Sunday 4 was prone to local extinction in almost every simulation. When the metapopulation does not include this population, chance of quasi-extinction appears to be reduced accordingly. We know, however, that the presumed high quality habitat of Sunday 4 and its central location in this linearly arranged chain of patches, makes this patch of central importance to bog lemming metapopulation on Sunday Creek. The relative importance of different habitat types should be examined for this species in order to quantify habitat quality for incorporation into a PVA using the Landscape Data subprogram in RAMAS/GIS. The program RAMAS/GIS could be a powerful tool for examining how different patches in a complex such as Suday Creek affect metapopulation process if density could reflect habitat quality. This leads to questions about what constitutes a viable population of northern bog lemmings. Three (somewhat) alternative hypotheses could apply: 1 ) lemmings live in habitat patches which have been isolated for thousands of years; 2) lemmings move substantial distances between patches supplementing (or recolonizing) the sub-population within a patch and contributing genetic material; and 3) lemmings use habitats other than moss bogs/fens. Alternative 1 . Populations within patches such as Wood Lake and Numa Ridge Bog would not appear to have been able to survive given the small habitat patch size, if they are indeed totally isolated and if lemmings do not use habitats other than moss mats. This leads us to think that this alternative is not completely feasible. Alternative 2. In several areas such as the Sunday Creek complex, the distribution and size of known patches suggests movement between patches. The overall view that most patches in Montana are relatively near other known, or potential, patches, gives support to this hypothesis. Arctic lemmings are known to make spectacular movements during highs in the population cycle; this could also be true of northem bog lemmings. Northern bog lemmings do undergo populations fluctuations at least in central Canada (Edwards 1 963). However, population cycles in general appear to be less dramatic in: 1 ) more southerly areas, and 2) in areas with less contiguous habitat for the cycling species. Alternative 3 . Lemmings have certainly been found in habitats other than bogs/fens in Montana and in other areas of their range. In the Montana sites where the habitat is atypical, captures represent a rare 16 event. Multiple trapping periods prior to and/or following a capture have not resulted in regular additional captures of lemmings. In Glacier National Park, general trapping for small mammals over nearly 1 00 years in numerous habitats has resulted in captures of 5 lemmings at 4 sites (all atypical habitats) (Wright 1 950, Hoffmann et al. 1 969, Weckwerth and Hawley 1 962, Pearson 1 99 1 ). In the rest of Montana, only 1 site has been found during general small mammal trapping (Shoofly Meadows, a typical habitat site) (Adelman 1 979). However, when trapping focused on bog/fen habitat, 1 2 new sites were discovered since 1990 (Pearson 1991, Reichel and Beckstrom 1993, 1994). Many of these sites have had multiple animals captured in a single night, supporting the premise that the fen\bog habitat is the primary habitat for northern bog lemmings in Montana. The extent of lemming use of other habitats has yet to be determined, but would appear to be low. Probably all three alternatives have some element of reality. It seems likely that 1) some patch complexes are isolated from others and have been for long periods of time; 2) some relatively long distance movements may increase gene flow, supplement small populations, and allow for recolonization of extirpated patches; and 3) while bog lemmings use a variety of habitats to a limited (and largely unknown) extent, bog and fen habitats hold the densest populations of lemmings. Research Methods. How do we get the information on distribution, habitat use, and movement that we need to manage this species? Distributional information, and to a lesser extent habitat use, has often been gathered using snap-traps. Detailed habitat use and movement data for small mammals are most commonly obtained using mark-recapture techniques with live traps. However, for northern bog lemmings, live traps are of very limited usefulness. This is because Sherman live-trap use: 1) is labor intensive throughout the trapping period; 2) has very low success with any bait tried; and 3) results in at least some mortality (4 of 6 known captures) (Pearson 1 99 1 , Reichel and Beckstrom 1 993). Pitfalls, used as live traps: 1 ) are labor intensive especially during placement; 2) cannot be used in the saturated soil situations commonly encountered in bog lemming habitat; and 3) result in at least some mortality during and between trapping periods. Given these drawbacks, it seems doubtful that live-trapping methods, by themselves, will yield much information on habitat use, population parameters, movements, or home range sizes. Incidental mortalities may be a significant factor over a study of sufficient duration to yield good information. Additionally, live-trapping to locate populations will require at least 1 times the effort and cost compared to snap-trapping, and still cause some mortality. Given the very low Sherman live-trapping success, negative results for even 1 000 trap-nights per site would not provide much confidence that lemmings are not present. Dropping boards may provide one option, but we think differentiating northern bog lemming droppings from other voles will be difficult. Jones and Bimey (1988) report that northern bog lemming droppings are bright green while other vole droppings are brown or black. However, we found that at least some bog lemmings had brown droppings. If color alone is used to differentiate the droppings, it may lead to serious biases. Pearson (1991) was not confident of identification of droppings {Microtus versus Synaptomys) in a test of the technique in Glacier National Park. He did speculate that it could be possible using more sophisticated identification techniques. 17 Snap-trapping for bog lemmings was much more successful than live-trapping although only 3 females were captured using this method (at all locations in Montana in 1 992 and 1 993). It appears to be the method of choice for initial survey work to find new populations, both from an economic and time- constraint view. Concerns have been expressed that snap-trapping is not a suitable technique to use on a "sensitive species." This argument may have some validity from a public perception point of view, but has little or no biological basis (Reichel and Beckstrom 1 993). Very small radio-telemetry packages have recently been used to study other voles and this technique seems to hold the most promise for studying Synaptomys. It would require relatively few individuals to be captured, and recapture of those individuals would not be necessary. It would seem to be the method of choice for examining activity patterns, habitat selection and use, home range size, and typical movements by Synaptomys. Long range movements, such as dispersal, are more difficult to determine using radio-telemetry. This is due to 1 ) the relative rarity of such movements; and 2) time and equipment limitations for finding animals moving far from their expected location. Indirect means of determining the amount of inter- patch movement are available using biochemical analyses of various types to measure gene fiow. This may be a viable approach to learning about inter-patch movements and gene fiow. STATEWIDE MANAGEMENT RECOMMENDATIONS AND RESEARCH NEEDS Based on limited observations at the sites where bog lemmings have been found, several interim management recommendations can be made. We feel that these are the minimum necessary to maintain viable bog lemming populations. Additional research is needed which may lead to other management actions necessary for maintaining viable bog lemming populations. 1) Lacking surveys at specific sites, assume northern bog lemmings are present at sphagnum or other fen/bog moss habitat patches in north Idaho and western Montana during land management planning processes. 2) Do not harvest timber within 1 00 m of sphagnum or other fen/bog moss mats or associated riparian areas which could provide corridors for inter-patch movements. 3) Minimize domestic livestock grazing in drainages with unsurveyed moss mats present. Range conditions in riparian areas with moss mats should be maintained in good to excellent categories. Stocking rates should be reduced to a point where rapid recovery occurs if either 1 ) current range condition is fair or poor; or 2) livestock are impacting moss mats. 4) No management activities which could destroy moss mats should be undertaken. Examples could include (but are not limited to): 1) road building in, or in some cases upslope from, bogs/fens; 2) pothole blasting in bogs/fens; 3) trail construction across or adjacent to bogs/fens; 4) dam construction upstream from bogs/fens, or downstream if fiooding of bogs/fens would occur; and 5) snowmobile use in bogs/fens which could compact vegetation or collapse lemming runways or nests. 18 Very little information is available on the northern bog lemming. Even the distribution in the U.S. is poorly understood, and most populations have been found within the past 1 5 years. Habitat use by northern bog lemmings has never been determined in any systematic way. Descriptions of occupied habitat consist of anecdotal accounts of where each specimen was captured; only about 35 individuals had been collected in the Pacific Northwest prior to 1 990. Reichel and Beckstrom (1 993, 1 994) contain detailed vegetative descriptions for six lemming sites in Montana. Food habits and reproductive information in the literature are also are limited to a very few anecdotal accounts. Analysis of food from stomachs of bog lemmings captured at six sites in western Montana show mosses composed 29-92% of the diet (by volume) with Sphagnum moss averaging <1%. Sedges (1-64%) and grasses (0-8%)) composed most of the rest of the diet (Reichel, unpubl. data). No information is available on such topics as movements, population densities, longevity, or home range. Much additional research is required to make intelligent land management decisions where northern bog lemmings are present. We recommend the following as the highest priority needs: 1) Conduct additional surveys to better understand macro- and micro- distribution in Montana; on a state-wide basis this should include surveys on the Dillon Resource Area, Headwaters Resource Area, Helena National Forest, Deerlodge National Forest, Gallatin National Forest, Custer National Forest, Lewis and Clark National Forest (Jefferson Division), and sites on the Beaverhead National Forest south and east of Maybee Meadows. 2) Analyze all stomachs of bog lemmings collected to provide additional food habits information; this should give some indication of potential habitat use. 3) Conduct plant community surveys at all known bog lemming locations. This should include identification of dominant mosses present. 4) Gather information on the autecological requirements of the mosses found at bog lemming sites. 5) Carry out research on northern bog lemming habitat use. Given the extreme difficulty in capturing the northern bog lemming, radio-telemetry is probably the only viable means to obtain satisfactory answers as to how bog lemmings use habitat within their home ranges. 6) Carry out research on northern bog lemming movements to gather information on home ranges and possibly dispersal. This information needs to be integrated with simultaneously collected habitat use data. Again, we feel radio-telemetry is the only viable methodology available. 7) Carry out biochemical research on allelic diversity and gene flow between habitat patches. It is possible that hair/skin from specimens already collected could be used for analysis. This should be done utilizing information on patch size and isolation, across the range of the lemming in Montana. Ideally, Montana information should be compared to information from a population in Canada at a site with relatively continuous habitat over a large area. 19 LITERATURE CITED Adelman, E. V. 1 979. A survey of the nongame mammals in the Upper Rattlesnake Creek drainage of Western Montana. M.S. Thesis, University of MT, Missoula. 129 pp. Akcakaya, H. R. 1 993 . RAMAS/GIS. Linking landscape data with population viability analysis. Applied Biomathematics, Seauket, New York. Anderson, R. M. 1932. Five new mammals from British Columbia. Natl. Mus. Can. Bull 70:99-107. Banfield, A. W. F. 1 974. The mammals of Canada. University of Toronto Press, Toronto. Beasley, L. E. and L. L. Getz. 1986. Comparison of demography of sympatric populations of Microtus ochrogaster and Synaptomys cooperi. Acta Theriologica 3 1 :3 85-400. Bursik, R. and R. K. Moseley. 1 992. Prospectus: Valley peatland ecosystem project, Idaho. Idaho Dept. Fish Game, Conservation Data Center, Boise, Idaho. Unpubl. Report 16 pp. Clough, G C. and J. J. Albright. 1 987. Occurrence of the northern bog lemming, Synaptomys borealis, in the northeastern United States. Canadian Field-Naturalist 101:611-613. Doutt, J. K., C. A. Heppenstall and J. E. Guilday. 1973. Mammals of Pennsylvania, 3rd edition. Penn. Game Comm., Harrisburg. 283 pp. Edwards, R. L. 1963. Observations on the small mammals of the southeastern shore of Hudson Bay. Can. Field-Nat. 77:1-12. Gaines, M. S., R. K. Rose, and L. R. McClenaghan, Jr. 1977. The demography oi Synaptomys C00/7OT in eastern Kansas. Can. J. Zool. 55:1584-1594. Gaines, M. S., C. L. Baker, and A. M. Vivas. 1979. Demographic attributes of dispersing southern bog lemmings {Synaptomys cooperi) in eastern Kansas. Oecologia 40:91-101 . Groves, C. and E. Yensen. 1 989. Rediscovery of the northern bog lemming {Synaptomys borealis) in Idaho. Northw. Nat. 70:14-15. Hall, E. R. 1 98 1 . The mammals of North America, 2nd edition, 2 vols., John Wiley and Sons, New York. 20 Hoffmann, R. S., P. L. Wright and F. E. Newby. 1 969. The distribution of some mammals in Montana I. Mammals other than bats. J. Mammal. 50:579-604. Johnson, M. L. and P. W. Cheney. 1953. Synaptomys in Idaho and northeastern Washington. Murrelet 34:10. Jones, J. K., Jr. and E. C. Bimey. 1988. Handbook of mammals of the north-central states. U. Minn. Press, Minneapolis. Jonkel, C. J. 1 959. An ecological and physiological study of pine marten. M.S. Thesis, Montana State Univ., Missoula. 81 pp. Layser, E. F. and T. E. Burke. 1 973 . The northern bog lemming and its unique habitat in northeastern Washington. Murrelet 54:7-8. Linzey, A. V. 1 983 . Synaptomys cooperi. Mammalian Species Accounts, No. 2 1 0. American Society of Mammalogists. 5 pp. Moseley, R. and C. Groves. 1 990. Rare, threatened and endangered plants and animals of Idaho. Unpubl. rep., Nat. Heritage Sect., Nongame and Endangered Wildl. Prog., Idaho Dept. Fish Game, Boise. 33 pp. Pearson, D. E. 1 99 1 . The northern bog lemming in Montana and the contiguous United States: distribution, ecology and relic species theory. Unpubl. Senior Thesis, Univ. Mont., Missoula. 33 pp. Reichel, J. D. 1 984. Ecology of Pacific Northwest alpine mammals. Ph.D. thesis, Washington State Univ., Pullman. 91 pp. Reichel, J. D. 1995. Montana animal species of special concern. [Unpubl. list], Mont. Nat. Heritage Prog., Helena. 10 pp. Reichel, J. D. and S. G Beckstrom. 1993. Northern bog lemmings survey: 1992. Montana Natural Heritage Program. Helena, MT. 64 pp. Reichel, J. D. and S. G Beckstrom. 1994. Northern bog lemmings survey: 1993. Montana Natural Heritage Program. Helena, MT. 87 pp. Shaffer, M. L. 1 98 1 . Minimum population sizes for species conservation. Bioscience 31:131-134. Weckwerth, R. P. and V. D. Hawley. 1 962. Marten food habits and population fluctuations in Montana. J. Wildl. Manage. 26(l):55-74. 21 Wilson, C, R. E. Johnson and J. D. Reichel. 1 980. New records for the northern bog lemming in Washington. Murrelet 6 1 : 1 04- 1 06. Wright, R L. 1950. Synaptomys borealis from Glacier National Park, Montana. J. Mammal. 3 1 :460. 22 APPENDIX I. NORTHERN BOG LEMMING PVA MODEL DESIGN PARAMETERS Life-cycle of the northern bog lemming. STAGE CLASSES Subadult Adult-first litter Adult-second litter APPROXIMATE AGES 90-1 80 days 180-270 days >270 days DURATION OF STAGE 3 months 3 months NA Population projection matrix for northern bog lemmings with a pre-birth pulse. ^380 §2 s. where: m^ = maternity rate subadults m^ = maternity rate adults first litter m3 = maternity rate adults second litter Sq = survival to subadult Sj = survival to adult S^ = survival of adults to breed a second time S3 = The projection interval is 3 months, the average interbirth interval. Calculation of variables in the projection matrix: m^ = maternity rate subadults = (average litter size of subadults)(proportion of subadult females breeding). We calculated the maternity rate of subadults as an average of the two projection intervals calculated from museum records: m^ = [(3.5)(0.57) + (3.25)(0.125)]/2 = 1.201 m^ = matemity rate adults first litter = (average litter size of adults first litter)(proportion of females breeding) 23 Calculated from museum records for females collected: m2=(4.28)(0.75) = 3.21 m3 = maternity rate adults second litter = (average litter size)(proportion of females breeding) Calculated from museum records for females collected: m3 = (3.9)(0.4848)= 1.89072 Sq = survival to subadult From Beasley and Getz (1 986) for southern bog lemmings in southern Illinois (see results): 8^ = 0.4100 Sj = survival of subadults to adult From Beasley and Getz (1 986) for southern bog lemmings in southern Illinois (see results): 8^ = 0.4100 S^ = survival of adults to breed a second time From Gaines et al. (1 997) for southern bog lemmings in Kansas: average adult 2-week survival rate in summer = 0.7067 (SE = 0.499) average adult 2-week survival rate in winter = 0.7683 (SE = 0.0601) Survival over the prediction interval (3 months), is the 2-week survival rate times 6, or 0.7067^ (=0. 1245) in summer, and 0.7683^ (=0.2057) in winter. Dispersing adults were taken into account by adding 0. 1 to summer and winter adult survival rates, increasing them to 0.2245 and 0.3057, respectively. Adults may breed a second time in the same summer they had their first litter, or may breed a second time after overwinter survival. Survival rate of adults to breed a second time was calculated as summer and winter survival rates weighted by maternity rates of first and second litters, respectively: S^ = (0.75)(0.2245) + (0.4848)(0.3057) = 0.1684 + 0.1482 = 0.3166 S3 = These variables were inserted in the Leslie matrix as follows: (1.201X0.4100) (3.21)(0.4100) (1.8907X0.4100) 0.4100 0.3166 24 The Leslie matrix calculated from the variables is: 0.4924 1.3161 0.7752 0.4100 0.3166 Demographic and environmental stochasticity Demographic and environmental stochasticity was modeled from the standard errors of the estimates for survival, litter size, and proportion of females breeding used in the population matrix. The stochasticity matrix takes the same form as the Leslie matrix for population projection, except in this case the variables represent the standard errors of the estimates rather than the average value for the variable: ([se]mJse]S„) ([se]mJse]S„) ([se]m3 [se]S„) (se)Sj (se)S2 (se)S3 Standard errors for survival were calculated from replicate samples (years and/or populations) reported in the literature. Litter size standard errors were from replicate females used to estimate the mean litter size, as presented in the Results section. These variables were inserted in the Leslie matrix as follows: (0.6915X0.2399) (1.63)(0.2399) (1.1)(0.2399) 0.2399 0.0419 The Leslie matrix for stochasticity is thus: 0.1659 0.3910 0.2639 0.2399 0.0419 Sources used to develop the model include: Akcakaya, H. R. 1 993 . RAMAS/GIS. Linking landscape data with population viability analysis. Applied Biomathematics, Seauket, New York. Brault, S., S. Boyd, F. Cooke, and J. Takekawa. n.d. Population models as tools for research cooperation and management: The Wrangel Island snow geese, unpubl. ms. 25 Caswell, H. 1989. Matrix population models. Sinauer Associates, Sunderland, MA. Mills, L. S., S. G. Hayes, C. Baldwin, M. J. Wisdom, J. Citta, D. J. Mattson, and K. Murphy. 1996. Factors leading to different viability predictions for a grizzly bear data set. Conservation Biology 10: 863-873. Wisdom, M. J. and L. S. Mills. 1 997. Sensitivity analysis to guide population recovery: prairie-chickens as an example. J. Wildl. Manage. 61 :302-3 12. 26 APPENDIX 11. MUSEUMS CONTACTED AND RESULTS Provincial Museum of Alberta, Edmonton: received data on 1 5 specimens University of Alberta Museum: received data on 27 specimens British Columbia Provincial Museum: received data on 77 specimens University of British Columbia: received data on 73 specimens Manitoba Museum of Man and Nature, Winnipeg: received data on 78 specimens University of Manitoba: no reply New Brunswick Museum, Saint John: received data on 1 specimen CarletonUniv. Mus. (Ontario): no reply Canadian Museum of Man and Nature, Ottawa: received data on 2 1 6 specimens Royal Ontario Museum, Toronto: received data on 147 specimens Redpath Museum, McGill University, QB: : have no S. borealis Saskatchewan Museum of Natural History: no reply University of Saskatchewan, Saskatoon: received data on 15 specimen AK Dept Fish and Game: called to say now specimens were primarily at Univ. AK, with some at TX A&MandUnivIU. University of Alaska Museum: received data on 74 specimens University of Arizona: : no reply California Academy of Sciences, San Francisco: have no S. borealis California State University, Fresno: : no reply Natural History Museum of Los Angeles County: have no S. borealis (phone 5/1 8/94) Museum of Vertebrate Zoology, University of California, Berkeley: received data on 39 specimens University of California, Los Angeles: received data on 1 specimens University of Colorado, Boulder: received data on 2 specimens University of Connecticut, Storrs: have no S. borealis National Museum of Natural History, Washington, DC: received data on 1 65 specimens Florida Museum of Natural History, University of Florida, Gainesville: have no S. borealis Field Museum of Natural History, Chicago, IL: received data on 5 1 specimens University of Illinois, Urbana: received data on 6 specimens Fort Hays State University, Museum of the High Plains, KS: : no reply University of Kansas, Lawrence: received data on 1 8 specimens Harvard University, Museum of Comparative Zoology, Cambridge: received data on 34 specimens Mich. State University, East Lansing: received data on 2 specimens University of Michigan, Ann Arbor: received data on 4 specimens University of New Mexico, Museum of Southwestern Biology: received data on 1 specimen American Museum Natural History, New York, NY: received data on 7 1 specimens University of Oklahoma: received data on 39 specimens Carnegie Museum of Natural History, Penn.: no reply Philadelphia Academy of Natural Sciences: have no S. borealis Texas A&M: no reply Texas Tech University, Lubbock: have no S. borealis 27