Use of Habitat by Mountain Beaver in the Sierra Nevada
Transcription
Use of Habitat by Mountain Beaver in the Sierra Nevada
J. Wildl. Manage. 53(3):1989 BROWSE IN SPRUCE-FIR FORESTS * Newton et al. Minn. Agric. Exp. Stn. Tech. Bull. 297. For. Ser. 15. 75pp. 649 browse on a big-game winter range in northern Idaho. J. Wildl. Manage. 20:141-151. , ANDH. L. HANSEN. 1969. Increasing browse NEWTON, M., K. M. HOWARD, B. R. KELPSAS, R. for deer by aerial applicationsof 2,4-D. J. Wildl. Manage. 33:784-790. W. C. 1981. How a forestaffectsa forage KRUEGER, crop. Rangelands3:70-71. DANHAUS, M. LOTTMAN, AND S. DUBELMAN. R. A. LAUTENSCHLAGER, 1985. Forestry, herbi- cides, and wildlife. Pages 299-307 in J. A. Bissonette, ed. Is good forestry good wildlife management? Me. Agric. Exp. Stn. Misc. Publ. 689. LEWIS,C. E., B. F. SWINDEL, L. F. CONDE, AND J. E. SMITH.1984. Forage yields improvedby site preparationin pine flatwoodsof north Florida. South.J. Appl. For. 8:181-185. MCCAFFERY, K. R., L. D. MARTOGLIO, AND F. L. 1984. Fate of glyphosate in an Oregon forest ecosystem.J. Agric. Food Chem. 32:1144-1155. , ANDF. N. DOST. 1984. Biological and phys- ical effectsof forestvegetationmanagement.Wa. Dep. Nat. Resour.Final Rep. Olympia. 424 pp. , AND L. A. NORRIS. 1968. Herbicide resi- dues in blacktail deer from forests treated with 2,4,5-T and atrazine.Pages 32-34 in Proc. West. Weed Contam. Conf., Boise, Id. SCIFRES, C. J., AND B. H. KOERTH. 1986. Habitat alterationsin mixed brushfrom variablerate herbicide patterns.Wildl. Soc. Bull. 14:345-356. JOHNSON. 1974. Maintaining wildlife openings VICARY, B. P., T. B. BRANN, AND R. H. GRIFFIN. with picloram pellets. Wildl. Soc. Bull. 2:40-45. 1984. Base-ageinvariantpolymorphicsite index curves for even-aged spruce-firstands in Maine. Me. Agric. Exp. Stn. Bull. 802. WALSTAD, J. D., ANDF. N. DOST. 1984. The health risks of herbicides in forestry: a review of the scientific record. Oregon State Univ. For. Res. Lab. Spec. Publ. 10. Corvallis.60pp. MCCORMACK, M. L., JR. 1982. Silviculture pro- gram. Pages 3-8 in Annu. Rep. Coop. For. Res. Unit, Sch. For. Resour.,Univ. Maine, Orono. . 1986. Vegetationproblemsand solutionsnortheast.South. Weed Sci. Soc. 38:315-326. , ANDM. NEWTON. 1980. Aerial applications of triclopyr,phenoxys,picloram,and glyphosate for conifer release in spruce-firforestsof Maine. Weed Sci. Soc. Am. 20:47-48. MUEGGLER,W. F. 1966. Herbicide treatment of Received 14 June 1988. Accepted 4 January 1989. BEAVERIN THESIERRANEVADA USE OF HABITATBY MOUNTAIN PAULBEIER,Departmentof Forestryand Resource Management,Universityof California,Berkeley,CA 94720 Abstract: I used stepwise logistic regressionto identify factors importantfor habitat use by SierraNevada mountainbeaver (Aplodontia rufa californica). Elevation, stream gradient, and abundanceof willow (Salix spp.), alder (Alnus spp.), and fir (Abies concolor and A. magnifica) had the strongest positive associations with habitat use. Mountainbeaver probably did not respond to these factors directly, but rather to a cool thermal regime, adequate soil drainage, and abundant food supply. Comparisonof current with historical beaver use indicated no marked change since 1935. J. WILDL. MANAGE.53(3):649-654 The mountain beaver is found in moist environments with moderate to dense vegetation in the Pacific Northwest and the Sierra-Cascades (Ingles 1965). The species is restricted to moist habitats because it has a poor ability to concentrate urine and requires free surface water or succulent vegetation on a daily basis (Nungesser and Pfeiffer 1964, Schmidt-Nielsen and Pfeiffer 1970). Although mountain beaver find appropriate vegetation or water on open slopes and ridgetops in the forests of western Oregon and Washington (Voth 1968, Kinney 1971), they are restricted to riparian areas in the relatively xeric Sierra Nevada (Stephens 1906, Orr 1949). However, not all areas with succulent vegetation are occupied, and no quantitative work has related occurrence of mountain beaver to physical and vegetative characteristics of their habitats. My objectives were to develop a model to identify the physical and vegetative variables associated with habitat use by mountain beaver in the Sierra Nevada, to validate the model on a set of stream reaches not used in development of the model, and to use historical records to determine whether habitat use has changed since 1935. J. S. Slaymaker assisted on most of the fieldwork. J. L. Patton and B. R. Stein provided access to the collections and field notes of the Museum of Vertebrate Zoology at the University 650 MOUNTAINBEAVERHABITATUSE * Beier of California, Berkeley. H. C. Black, J. E. Borrecco, J. Evans, D. R. McCullough, S. H. Jenkins, and 2 anonymous reviewers offered helpful criticisms of the paper. This research was supported by the California Department of Fish and Game and the California Agricultural Experiment Station Project 4326-MS. STUDY AREA The study was conducted on the Truckee River and its tributaries from the confluence with Deer Creek (Nevada County, Calif.) downstream to Verdi (Washoe County, Nev.), a drainage of approximately 600 km2 with 172 km of streams ranging from 1,485 to 2,750 m in elevation. Stream banks were characterized by deciduous riparian vegetation that included aspen (Populus tremuloides), black cottonwood (P. trichocarpa), willows, alder, dogwood (Cornus spp.), bitter cherry (Prunus emarginata), lodgepole pine (Pinus contorta), and herbaceous plants. The vegetation of the area was described in detail by Beier and Barrett (1987). METHODS Each stream was divided into sampling reaches (approx 700 m long). Length varied from 400 to 900 m to put the end points at topographic or vegetative discontinuities and minimize heterogeneity within the sampling unit. For each reach, 7 physical and 12 vegetation variables (Table 1, plus aspect) were measured between 5 May and 20 August 1985. I measured bank slope class, stream width, and stream depth class near the midpoint of the reach, avoiding atypical locations such as waterfalls and beaver impoundments. Elevation, aspect, and gradient were obtained from 1:24,000 topographic maps; aspect was categorized into 1 of the 8 cardinal directions, or as flat (i.e., slope <2%). I measured bare soil, litter cover, riparian zone width, and abundance of herbaceous plants by taking the average of ocular estimates at 100-m intervals within each reach. The other vegetation variables were obtained by counts, except when it was obvious that the species abundance would be scored "3" (i.e., >20 plants/100 m of stream length). Mountain beaver presence was based on occurrence of the burrows and haypiles characteristic of the species, in association with bevel-cut woody stems (10-14 mm in diam). Beaver (Castor canadensis) activity in each reach was categorized as active colony; abandoned colony; beaver foraging, but no past or present J. Wildl. Manage. 53(3):1989 colony; or no sign of beaver. Two beaver-use variables were used in analyses. One variable was dichotomous(beaverpresentor absent),and the second was polychotomous(4 categories of beaver activity). Most habitat variables were measured on a scale with only 4 possiblestates,and other variables (aspect and beaver use) were categorical; such variablescannot be normally distributed. The continuousvariableswere not normallydistributed(Kolmogorov-Smirnovtests, P < 0.01); gradients,for instance, had a strongly bimodal distributionwith peaks at 0 and 10%,and were skewed to the right. Box'sM-statistic(Nie 1983) indicated that group variances were heterogeneous (P < 0.01), even after logarithmic or square-root transformations.Stepwise logistic regression (SLR) is a classification procedure well-suitedfor use with polychotomousand discontinuous explanatory variables (Walker and Duncan 1967, Cox 1970)and is appropriateeven when the conditions of normality and homogeneous variancesare not met (Anderson1972, Press and Wilson 1978). Therefore, SLR was used to identify variablesassociatedwith habitat use by mountain beaver. Variables that differed (P < 0.05) between used and unused reaches were candidates for entry into the logistic function. At each step, the variable with the lowest P-value was included until the significanceof the F-statisticof each remainingvariablewas >0.05. I randomly selected half the cases for use in model development; the other half were used for model validation.Priorprobabilitieswere based on the proportionof cases falling into each group. The ratioof each coefficientto its standarderrorwas used to standardizecoefficient size to facilitate comparisonof variables included in the function. I contrasted the frequency distributionsof gradientsand elevations at sites used by mountain beaver in the Truckee River Basin in 1985 to the gradients and elevations at pre-1935 mountain beaver sites in the Sierra Nevada (basedon publishedrecords,collections,and field notes of the Museum of Vertebrate Zoology [MVZ],Univ. California, Berkeley) to investigate change in habitat use over time. RESULTS Habitat Use Mountainbeaver were present on 56 (24%) of the 237 reaches surveyed. The reaches on MOUNTAIN BEAVER HABITAT USE * Beier J. Wildl. Manage. 53(3):1989 651 Table 1. Meanvalues for 18 habitatvariableson reaches used (n = 56) versus not used (n = 181) by mountainbeaverin the TruckeeRiverBasin,Californiaand Nevada, 1985. Significantdifferencesbetween means are denotedby an asterisk(t-test, P < 0.01). The last 3 columns show statistics for variablesincludedin the logistic function,which was developed using 118 randomlyselected reaches. Mountain beaver Variable Present 13.6 1.9 5.2 Stream gradient (%) Bank slope classt' Bare soil (%) Logistic function Absent * 5.0 1.6 6.0 17.3 4.4 * Stream depth class' 1.6 * 2.6 Elevation (m) Riparian zone width 2,137 19.4 * 1,823 26.9 Stream width (m) Litter cover (%) Aspen,' Cottonwood,' Alder, Willow' Dogwood' Bitter cherry"' Lodgepole pine' Yellow pine,' Fir,' Herbaceous plants, Constant (m) 4.5 1.1 0.3 2.1 2.4 1.3 0.8 0.9 0.6 1.6 2.3 * * * * * * 3.2 0.5 0.9 1.7 2.0 1.0 1.2 0.9 1.1 0.9 2.3 Ratioa Tolerance 0.244 3.28 0.63 0.0069 3.67 0.70 0.885 1.396 2.39 2.78 0.86 0.95 1.395 2.24 0.59 Coefficient -25.76 -4.09 a Ratio of coefficient to its SE. b Slope class: =1 1-30, 2 = 31-40, 3 = 41-60, 4 = >60%. c Depth class: 1 1-23, 2 = 24-46, 3 = 47-76, 4 = >76 cm. (1 Abundance index: 0 = absent, 1 = 1-5, 2 = 6-20, 3 = >20 trees or shrubs in riparian zone/100 m of stream length. e Abundance index: 0 = absent, 1 = rare, 2 = moderate, 3 = abundant. which mountain beaver were present differed significantly (P < 0.01) from the unused reaches with respect to 10 of 18 non-categorical habitat variables (Table 1). Of the categorical variables, there was significantly less beaver activity on reaches used versus not used by mountain beaver (x2 = 23.6, 3 df, P < 0.01), but aspect was not associated with mountain beaver presence (x2 = 5.9, 8 df, P > 0.40). On average, reaches with mountain beaver had steeper gradients, narrower and shallower streams, were higher in elevation, had a greater abundance of alder, willow, fir, and aspen, and a lesser abundance of cottonwood and yellow pine (Pinus ponderosa, P. jeffreyi, P. washoensis, and hybrids) than unused reaches. Gradients and elevations differed most strikingly between used and unused reaches (Fig. 1). The logistic regression model, using 118 randomly selected reaches, included these 2 variables and abundances of alder, willow, and fir (Table 1). When applied to the 119 cases not used to develop the model, the logistic function correctly classified 65% of the used reaches and 97% of the unused reaches. Although 60% of 237 reaches showed sign of past or present use by beaver, only 16 (29%) of 56 reaches used by mountain beaver also had sign of beaver use and none of these 56 reaches contained active beaver colonies. However, SLR did not select either of the 2 beaver-use variables, and the forced addition of a beaver-use variable into the function did not improve the classification. HistoricalPatterns of HabitatUse The collections and field notes at the University of California's MVZ provided 29 mountain beaver sites in the Sierra Nevada between 1912 and 1935 for which elevation and gradient could be determined. These records indicate no change in habitat use by mountain beaver. Only 4 of the 29 sites were < 1,700 m, whereas 13 were >2,300 m in elevation. No site had a stream gradient <5% and only 3 sites had gradients between 5 and 9%, whereas 13 had gradients >19%. Collectors C. L. Camp and T. I. Storer provided 12 of the 29 records, and searched extensively for mountain beaver in the Yosemite region during 1915-19. Their field notes suggest that mountain beaver were not widely distributed at that time. I ....... MOUNTAINBEAVERHABITATUSE * Beier 652 A J. Wildl. Manage. 53(3):1989 the ground for >4.5 hours/day (Ingles 1959, Voth 1968), and high ambient temperatures 56 USEDREACHES probably exclude the species from low eleva181 UNUSED REACHES 50tions in the Sierra Nevada. It is not obvious why steep gradients would be preferred by mountain beaver. Gradients 30were not highly correlated with the other explanatory variables (r2 = 0.26 with elevation, pl 00< less with other variables). Gradient may be imKXK C> 10OCK OCK portant because it promotes water drainage, thus 00< CO) w nrr7iF% Ix preventing burrows from flooding. This interLI 0-4 5-9 10-14 15-19 pretation is supported by observations on 3 0 reaches used by mountain beaver in which there GRADIENT (%) was a break in slope from steep (> 10%) to gentle B (<5%) gradient, with no noticeable change in 56 USEDREACHES LL bank slope, or other variables. In vegetation, 181 UNUSED REACHES 0 such cases, mountain beaver were found only 50in the steeper part of the reach. Crouch (1964) found greater numbers of mountain beaver on west and northwest facing 30slopes, an observation consistent with the importance of cool temperatures, and with the fact that moisture-bearing storms come from these directions. Although this trend was not evident 10in the Truckee River Basin, only 19 of the 237 LXXX A AAA reaches had southeast to southwest aspects. This < 1,700 1,700-2,000 2,000-2,300 > 2,300 may have been too small a sample to detect ELEVATION aversion to these warmer slopes. (M) variables were also important, alVegetation 1. Relative distributions for Fig. frequency gradients(A)and elevations(B) on reaches used versus not used by mountain though they contributed less to the logistic funcbeaver in the Truckee River Basin, Californiaand Nevada, tion. Willow, alder, and fir are important foods 1985. (P. Beier, unpubl. data) and also provide thermal and escape cover. In Washington and Oregon, some sites used by mountain beaver had DISCUSSION very little alder and willow, but were abundant HabitatUse in ferns, which formed the bulk of the animal's Elevation was the most important factor re- diet (Voth 1968, Allen 1969). These data suggest lated to habitat use by mountain beaver. Ste- that although mountain beaver require an adphens (1906) also described mountain beaver equate supply of succulent vegetation, including habitat in the Sierra Nevada as high elevation evergreen winter food, the species composition hillsides, but in western Oregon, the species is of the vegetation can be variable. I suggest that abundant at sea level (Voth 1968, Johnson 1971, habitat use by mountain beaver involves strict Kinney 1971). In the Sierra Nevada, high elerequirements for an appropriate thermal revation was probably preferred because it is as- gime and adequate soil drainage, and somewhat sociated with lower mean temperatures. The more flexible requirements for food. ability of a fossorial mammal to tolerate warm Although mountain beaver and beaver did temperatures decreases as body mass increases not co-occur in the Truckee River Basin, beaver (McNab 1979). This relationship may play a role variables did not improve the classification funcin the distribution of mountain beaver, 1 of the tions. Thus, the lack of spatial overlap between largest fossorial mammals. In laboratory exper- the 2 species in the Sierra Nevada is probably iments, mountain beaver were unable to sustain due to different habitat requirements rather than activity at temperatures >28 C (Johnson 1971, to displacement. This conclusion is supported Kinney 1971). Mountain beaver forage above by observations of the 2 species coexisting at *i* I ,.A. J. Wildl. Manage. 53(3):1989 many sites in Washington and Oregon (J. E. Borrecco, U.S. For Serv. and J. Evans, U.S. Fish Wildl. Serv., pers. commun.). Historical Patterns of Habitat Use In their pre-1935 descriptions of mountain beaver habitat in the Sierra Nevada, Camp (1918) and Grinnell and Storer (1924) did not comment on gradient or elevation. However, Stephens (1906:94) found mountain beaver in the Sierra Nevada "in canyons and on mountain sides where suitable springs occur, usually at considerable altitude." Historical records indicate that the mountain beaver's association with high elevations and steep gradients is not the result of human activities excluding the animal from low elevation, low gradient sites since 1935. The field notes of C. L. Camp and T. I. Storer also suggest that suitable habitat has been scarce in the Sierra Nevada. T. I. Storer (1919 field notes:829) commented on their rarity: "It would seem that the hold which these animals have in the scheme of existence is exceedingly slight, being so especially fitted for a particular kind of life and a particular sort of habitat." Because I could find no pre-1906 references to mountain beaver in the Sierra Nevada, I could not detect any shifts in habitat use by mountain beaver prior to the 1906-35 era. MANAGEMENTIMPLICATIONS In the Sierra Nevada, mountain beaver occur in small patches of high elevation, steep, moist habitat, often separated from other populations by distance and topography. The historical record indicates that this is a long-standing pattern, because the Sierra Nevada apparently offer only marginal habitat conditions. Multivariate wildlife habitat models can be used for 2 different purposes, namely to identify factors important in habitat use, and to predict species occurrence (Shugart 1981). With respect to the first goal, the SLR model suggests that a cool thermal regime and adequate soil drainage are the most important variables governing habitat use by mountain beaver in the Sierra Nevada, and that an adequate supply of succulent food, including some evergreen foliage, is also necessary. Therefore, management activities such as road building, livestock grazing, and herbicide applications will also influence suita- MOUNTAINBEAVERHABITATUSE * Beier 653 bility of the habitat for mountain beaver, by altering soil drainage and forage composition. As a predictor of mountain beaver occurrence, the SLR model correctly identified 97% of the reaches not used by mountain beaver, and 65% of the used reaches. From a management perspective, the latter rate is more critical, because the failure to identify used habitats is a more serious error than incorrectly predicting the presence of a population (Rice et al. 1981). Because the model may frequently fail to alert the manager to potential impact on the species, it should not be used to predict mountain beaver occurrence in the Sierra Nevada. LITERATURECITED ALLEN,L. 0. 1969. Preferential food habits of Aplodontia rufa, M.Ed. Thesis, Central Washington State Coll., Ellensburg.55pp. ANDERSON,J. A. 1972. Separate sample logistic discrimination. Biometrika 59:19-35. BEIER,P., ANDR. H. BARRETT.1987. Beaver habitat use and impact in the Truckee Basin, California. J. Wildl. Manage. 51:794-799. CAMP, C. L. 1918. Excavations of burrows of the rodent Aplodontia, with observations on the habits of the animal. Univ. California Publ. Zool. 17: 517-536. Cox, D. R. 1970. The analysis of binary data. Methuen and Co. Ltd., London, U.K. 142pp. CROUCH, G. L. 1964. Forage production and utilization in relation to deer browsing of Douglasfir seedlings in the Tillamook burn, Oregon. Ph.D. Thesis, Oregon State Univ., Corvallis. 162pp. GRINNELL, J., AND T. I. STORER. 1924. Animal life in the Yosemite. Univ. California Press, Berkeley. 752pp. INGLES,L. G. 1959. A quantitative study of mountain beaver activity. Am. Midl. Nat. 61:419-423. 1965. Mammals of the Pacific states. Stanford Univ. Press, Stanford, Calif. 506pp. JOHNSON, S. R. 1971. Thermal regulation, microclimate, and distribution of the mountain beaver, Aplodontia rufa pacifica Merriam. Ph.D. Thesis, Oregon State Univ., Corvallis. 164pp. KINNEY, J. L. 1971. Environmental physiology of a primitive rodent. Ph.D. Thesis, Univ. Oregon, Eugene. 181pp. McNAB, B. K. 1979. The influence of body size on the energetics and distribution of fossorial and burrowing mammals. Ecology 60:1010-1021. NIE, N. H., editor. 1983. SPSSX user's guide. McGraw-Hill,New York,N.Y. 806pp. 1965. Water balance and maximum concentrating capacity in the primitive rodent, Aplodontia rufa. Comp. Biochem. Physiol. 14:289-297. ORR, R. T. 1949. Mammals of Lake Tahoe. Calif. Acad. Sci., San Francisco. 127pp. PRESS, S. J., AND S. WILSON. 1978. Choosing beNUNGESSER, W. C., AND E. W. PFEIFFER. 654 MOUNTAINBEAVERHABITATUSE * Beier tween logistic regressionand discriminantanalysis. J. Am. Stat. Assoc. 73:699-707. RICE, J., R. D. OHMART, AND B. ANDERSON. 1981. Bird community use of riparianhabitats. Pages 186-196 in D. E. Capen, ed. The use of multivariatestatisticsin studiesof wildlife habitat.U.S. For. Serv. Gen. Tech. Rep. RM-87. SCHMIDT-NIELSEN, B., AND E. W. PFEIFFER. 1970. Urea and urinary concentrating ability in the mountainbeaver, Aplodontia rufa. Am. J. Physiol. 218:1370-1375. SHUGART,H. H., JR. 1981. An overview of multi- variate methods and their application to studies of wildlife habitat. Pages 4-10 in D. E. Capen, ed. The use of multivariatestatisticsin studiesof J. Wildl. Manage. 53(3):1989 wildlife habitat. U.S. For. Serv. Gen. Tech. Rep. RM-87. F. 1906. Californiamammals.WestCoast STEPHENS, Publ. Co., San Diego, Calif. 351pp. VOTH,E. H. 1968. Food habitsof the Pacificmountain beaver, Aplodontia rufa pacifica Merriam. Ph.D. Thesis, Oregon State Univ., Corvallis. 263pp. WALKER,S. H., AND D. B. DUNCAN. 1967. Estimation of the probabilityof an event as a function of severalindependentvariables.Biometrika 54:167-179. Received 2 October 1987. Accepted 4 January 1989. PREDATION ON UNIONIDCLAMSBY MUSKRATS SIZE-SELECTIVE LAURIEE. CONVEY,Departmentof Zoology,Universityof Alberta,Edmonton,AB T6G2E9, Canada JOHNMARKHANSON,Departmentof Zoology,Universityof Alberta,Edmonton,AB T6G2E9, Canada WILLIAM C. MACKAY, Departmentof Zoology, Universityof Alberta,Edmonton,AB T6G2E9, Canada Abstract: We collected shells of the northern floater (Anodonta grandis simpsoniana) from muskrat (Ondatra zibethicus) middens on the shore of a small lake in the boreal forest zone of Alberta. Muskratsate a mean of 228 ? 23.6 (SE) clams/day or 1.4 ? 0.15 kg/day (soft body mass measuredas wet wt) from 22 July to 1 September 1986. The overall length and age distributionsof clams eaten (median length = 64.3 mm, median age = 7.5 yr) were larger and older (P < 0.001) than a random sample of clams from the lake (median length = 49.1 mm, median age = 6.2 yr). Muskratsmay have a significant effect on the size and age distributionsof clams in the lake. MANAGE.53(3):654-657 J. WILDL. The muskrat is a herbivore that occasionally eats animal matter (Enders 1932, Smith 1938, Bellrose 1950). Most studies of muskrat diet have reported only the plant species in their diet (Butler 1940, Takos 1947, Danell 1978) because most studies of muskrat diets are conducted in marshes where plant matter is abundant relative to suitable animal prey. A few studies noted the consumption of animal prey (e.g., crayfish, fish, insects, snails, young birds, other muskrats, frogs, turtles, salt- and freshwater mussels) (Errington 1941, Bellrose 1950, Triplet 1983). In addition, muskrats have been accused of destroying mussel beds (Headlee 1906, Van Cleave 1940, Joy 1985) based on the observation of shells discarded on feeding platforms. However, prey selection and consumption rates of clams by muskrats have not been documented. The goals of our study were to determine the size and age distributions of clams consumed by muskrats, to test for spatial and temporal variation in their size and age distributions, and to determine whether the size and age distributions of clams eaten differed from the size and age distributions of clams in the lake. We thank S. A. Boutin for reviewing this paper. C. Podemski provided extensive field assistance. LEC was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Summer Undergraduate Award. JMH was supported by an NSERC Postdoctoral Fellowship. This study was done at the Meanook Biological Research Station and was funded by a Boreal Alberta Research grant to JMH and an NSERC operating grant to WCM. METHODS We collected intact valves of unionid clams from muskrat middens along the shore of Narrow Lake (54?34'N, 113?37'W), a small (1.14 km2), deep (x depth = 14.2 m), moderately productive (x total phosphorus = 12.9 mg/m2) lake