Integrative Zoology 2019; 14: 528–541 doi: 10.1111/1749-4877.12397

1 ORIGINAL ARTICLE 1 2 2 3 3 4 4 5 5 6 6 7 Impact of climate change on the small mammal community of the 7 8 8 9 Yukon boreal forest 9 10 10 11 11 12 12 13 Charles J. KREBS,1 Rudy BOONSTRA,2 B. Scott GILBERT,3 Alice J. KENNEY1 and Stan 13 14 14 4 15 BOUTIN 15 16 1Department of Zoology, University of British Columbia, Vancouver, British Columbia, Canada, 2Department of Biological 16 17 Sciences, University of Toronto Scarborough, Toronto, Ontario, Canada, 3Renewable Resources Management Program, Yukon 17 18 18 College, Whitehorse, Yukon, Canada and 4Department of Biological Sciences, University of Alberta, Edmonton, Alberta, Canada 19 19 20 20 21 21 22 Abstract 22 23 23 24 Long-term monitoring is critical to determine the stability and sustainability of wildlife populations, and if 24 25 change has occurred, why. We have followed population density changes in the small mammal community in 25 26 the boreal forest of the southern Yukon for 46 years with density estimates by live trapping on 3–5 unmanipulat- 26 27 ed grids in spring and autumn. This community consists of 10 species and was responsible for 9% of the ener- 27 28 gy flow in the herbivore component of this ecosystem from 1986 to 1996, but this increased to 38% from 2003 28 29 to 2014. Small mammals, although small in size, are large in the transfer of energy from plants to predators and 29 30 decomposers. Four species form the bulk of the biomass. There was a shift in the dominant species from the 30 31 1970s to the 2000s, with Myodes rutilus increasing in relative abundance by 22% and Peromyscus maniculatus 31 32 decreasing by 22%. From 2007 to 2018, Myodes comprised 63% of the catch, Peromyscus 20%, and Microtus 32 33 species 17%. Possible causes of these changes involve climate change, which is increasing primary production 33 34 in this boreal forest, and an associated increase in the abundance of 3 rodent predators, marten (Martes ameri- 34 35 cana), ermine (Mustela ermine) and coyotes (Canis latrans). Following and understanding these and potential 35 36 future changes will require long-term monitoring studies on a large scale to measure metapopulation dynamics. 36 37 The small mammal community in northern Canada is being affected by climate change and cannot remain sta- 37 38 ble. Changes will be critically dependent on food–web interactions that are species-specific. 38 39 Key words: community change, long-term study, population cycles, trophic dynamics, voles 39 40 40 41 41 42 42 43 43 44 INTRODUCTION 44 45 45 46 Much of traditional ecological theory is stabili- 46 47 ty-based, and the advent of climate change has forced 47 48 Correspondence: Charles Krebs, Department of Zoology, ecologists to consider the time scale of relative stability 48 49 University of British Columbia, 6270 University Blvd., in ecosystems. In this paper we report the small mam- 49 50 Vancouver, B.C. V6T 1Z4 Canada. mal community dynamics in a Yukon boreal forest over 50 51 Email: [email protected] a time period of 46 years. The boreal forest occupies ap- 51

528 © 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Climate change impact on a boreal forest community

1 proximately 57% of the Canadian land surface and is Boonstra and Krebs (2006) examined the demography 1 2 dominated by evergreen coniferous trees, typically white of Myodes at Kluane during the Kluane Ecosystem ex- 2 3 spruce (Picea glauca) in the forested valleys of the Klu- periments from 1986 to 1996 and suggested 3 hypothe- 3 4 ane Lake area. Common boreal forest trees like black ses regarding the causes of population changes in Myo- 4 5 spruce (Picea mariana) and lodgepole pine (Pinus con- des: predation, food and weather. With 20 more years of 5 6 torta) are absent in the Kluane Lake area but present in Myodes data we can test these hypotheses. 6 7 surrounding regions. The ground vegetation in the for- We wish to answer the following specific questions 7 8 ested area is comprised of an array of perennial plants for small rodents in the Yukon boreal forest: 8 9 9 of low diversity (Turkington et al. 2014). An extensive 1. Do populations of the common species fluctuate peri- 10 10 alpine zone exists above the forested valleys. The her- odically in 3–4-year cycles? 11 bivorous trophic level in the forested zone is dominated 11 2. Do populations of the different species fluctuate in 12 by snowshoe hares (Lepus americanus Erxleben, 1777), 12 phase? 13 which fluctuate in a 9–10-year cycle. This cycle affects 13 14 many, but not all, species in the food web and has been 3. Are these population fluctuations disappearing as a 14 15 reviewed extensively in Krebs et al. (2001, 2018a). We result of climate change? 15 16 concentrate here on the small mammal community, only 4. What ecological factors involving predation, food 16 17 one part of the boreal forest fauna described in Krebs supply and/or weather drive these demographic 17 18 et al. (2001) and in Boonstra et al. (2018). changes? 18 19 Beginning with Charles Elton (1942), much research 5. Does the dominant 9–10-year cycle of snowshoe 19 20 has focused on periodic fluctuations of small mammals hares affect small rodent population dynamics? 20 21 in the Northern Hemisphere, which were considered an 21 22 anomaly in the paradigm of the stability of nature. Much MATERIALS AND METHODS 22 23 recent work has concentrated on describing the demog- 23 24 raphy of these cycles, identifying populations that do Ten small rodent species occur in the Kluane region. 24 25 not fluctuate in a regular pattern, and trying to uncover Red-backed voles [Myodes rutilus (Pallas, 1779)] are 25 26 the demographic causes of population changes and the the most common, comprising approximately 70% of 26 27 limiting factors behind these changes (Krebs 2013). Of the biomass of this group. Deer mice [Peromyscus ma- 27 28 all the possible limiting factors, food shortage, preda- niculatus (Wagner, 1845)] and 4 species of voles (Mi- 28 29 tion, disease and social behavior have been most stud- crotus spp.) are less abundant. In addition, there are 4 29 30 ied (Boonstra & Krebs 2012; Radchuk et al. 2016). rare species present but rarely caught (Krebs & Wingate 30 31 Very few small mammal ecologists have concentrated 1976), as well as chipmunks (Tamias minimus Bach- 31 32 on weather as a direct limiting factor for rodent popula- man, 1839), and shrews on which we have inadequate 32 33 tions (the exception is Fuller 1969, 1977) on the implied data. Monitoring for changes in the abundance of small 33 34 assumption that weather must act through the more im- mammals in the Kluane region has been carried on an- 34 35 mediate factors of food shortage or predation. One con- nually since 1973. Many of the details of these methods 35 36 sequence of this omission is that if you ask what effect are described in Boonstra et al. (2001). From 1973 to 36 37 climate change might have on any particular rodent pop- 1975 we sampled widely throughout the Kluane valley 37 38 ulation, there is, at present, little insight. Attention be- with Museum Special snap traps and on 4 live-trapping 38 39 came focused in Europe on the role of climate change grids with Longworth live traps. Our early studies with 39 40 when evidence accumulated from long-term studies sug- snap trapping demonstrated that voles and mice in the 40 41 gested that populations were showing attenuated cycles, Kluane area showed spatial synchrony (Krebs & Win- 41 42 and that possibly small rodent cycles could disappear gate 1976, 1985). After 1975, virtually all our data came 42 43 (Cornulier et al. 2013). from live trapping standard grids (10 by 10 with 15-m 43 44 44 Most small mammal studies are short-term, of the or- spacing [2.3 ha] with either 50 or 100 Longworth live 45 45 der of 3–5 years, yet the importance of long-term stud- traps). We used a variety of sites in the area just south of 46 46 ies has been widely recognized (Likens 1989; Hughes Kluane Lake for live trapping in the first 10 years of our 47 47 et al. 2017). In this paper we report on 46 years of popu- studies, but by 1979 we had begun trapping Grid J, one 48 48 lation changes in small rodents of the Yukon boreal for- of our standard control grids for small rodents, and it 49 49 est. We use these studies to infer the patterns of change has been trapped continuously ever since. We set up live 50 50 and the potential limiting factors in these populations. trapping grids in 1973 at Mile 1050 of the Alaska High- 51 51

© 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, 529 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd C. J. Krebs et al.

1 way, and this grid was shifted slightly to become a stan- pact of bird predation on small rodents in this ecosys- 1 2 dard control grid (Silver) in 1987. We have put all the tem. Our studies on raptor predation extended only 2 3 available data together here from all our live trapping from 1987 to 1996 and are reported in Doyle and Smith 3 4 to obtain a comprehensive view of small mammal pop- (2001). The most extensive data available from these 4 5 ulation changes, fully realizing that the data are incom- studies was on the great-horned owl (Bubo virginianus 5 6 plete for several small mammal species that are present Gmelin, 1788). 6 7 but rare. There are 3 habitats that consist of over 90% of We have constructed energy flow estimates for the 7 8 the landscape of the Kluane region: closed spruce for- most common mammals and birds of the Kluane area 8 9 est, open spruce forest, and willow shrub. These are from equations for energy use given in Nagy et al. (1999) 9 10 intermingled at a fine scale and we found in our ear- and estimates of average density over 10 years from 2 10 11 ly work from 1973–1975 that all 3 habitats contained a periods, 1986–1996 and 2003–2014. We appreciate that 11 12 similar small mammal fauna. We set up sampling grids these energy pyramids are only approximate and rep- 12 13 at convenient locations in open and closed spruce for- resent a 10-year average, but they illustrate in an im- 13 14 est, which are the 2 most common habitats in the val- portant way how the terrestrial vertebrate community at 14 15 ley. All the 3 major habitats are a product of fires 150 to Kluane has changed over time. 15 16 16 300 years ago. We were not able to sample any of the Two small mammalian predators – marten [Mar- 17 17 rare habitats in the Kluane region, such as damp habitats tes americana (Turton, 1806)] and ermine (Mustela er- 18 18 along water courses and ponds, and we have not includ- minea Linnaeus, 1758) – are established in the Kluane 19 19 ed samples taken in the alpine above treeline. region but vary greatly in their abundance. Both spe- 20 20 Live-trapped rodents were typically sampled in 2-day cies are major predators of small rodents. Ermine were 21 21 sessions with 3 trap checks and whole oats & apple as caught in our live traps, but not frequently. A small num- 22 22 bait in Longworth live traps. We trapped after snow ber of marten (n = 30) were transplanted to the Kluane 23 23 melt in spring at the start of the breeding season (early area in 1984–1987 to augment the resident, low-densi- 24 24 May) and in early autumn (September) typically at the ty population (Slough 1989). While initial survival of 25 25 end of breeding. We used Efford’s spatially-explicit cap- transplanted marten was high, the success of this small- 26 26 ture–recapture program (SECR, Efford & Fewster 2013) scale effort in effectively augmenting the resident pop- 27 27 to estimate density except when the number of animals ulation is unknown. The least weasel (Mustela nivalis 28 28 captured was 4 or fewer when we used the minimum Linnaeus, 1766) has been caught in the Kluane area but 29 29 number known alive and the effective trapping area of is rarely present. One was captured by hand over the pe- 30 30 our grids (2.8 ha). Trappability was very high for all the riod 1986–1996, and we did not catch any in annual live 31 31 major rodent species. We calculated 2 indices of popu- trapping from 1986 to 2018. 32 lation change (summer growth and overwinter decline) 32 33 Estimates of terrestrial predator abundance in the 33 using live-trap data. Summer population growth was the Kluane area have been indexed by snow tracking 34 Fall estimate (t)/Spring estimate (t) and overwinter pop- 34 35 (O’Donoghue et al. 2001). These snow track transects 35 ulation decline was the Spring estimate (t + 1)/Fall esti- have been carried out every year since 1986 along the 36 mate (t). 36 37 same 25-km transect. Snow tracking data estimate ac- 37 To estimate the potential food supply for small ro- 38 tivity that we assume to be related in some approximate 38 dents we counted ground berries in the boreal forest at 39 way to population abundance. It is possible that ermine 39 10 locations every summer from 1995 to 2018. Five ma- 40 abundance is underestimated if they do not enter Long- 40 jor ground shrubs produce ground berries in the Kluane 41 worth traps often. Snow tracking in the autumn when 41 forest: bearberry (Arctostaphylos uva-ursi), red bearber- 42 minimal snow is present should give us a reliable index 42 ry (A. rubra), cranberry (Vaccinium vitis-idea), crow- 43 of abundance, but we cannot yet translate that to abso- 43 berry (Empetrum nigrum) and soapberry (Shepherdia 44 lute abundance. 44 canadensis). Toadflax (Geocaulon lividum) was also 45 Weather changes in the Kluane area have been re- 45 present in many locations. We monitored the production 46 corded by the Haines Junction Meteorological Station 46 of ground berries annually on an average of 707 40 × 47 of Environment Canada since 1945 and provide a gen- 47 40-cm fixed quadrats (Krebset al. 2009). 48 eral background for the observed biological changes. 48 49 We do not have data on the abundance of bird preda- In addition to the standard measurements of tempera- 49 50 tors for our sites for most of our monitoring years, and ture, rainfall and snow depth, since 2004 we have mea- 50 51 this is a large gap in our attempts to determine the im- sured temperature profiles in winter using I-buttons 51

530 © 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Climate change impact on a boreal forest community

1 (Maxim Thermochron iButtons DS 1921G-F5) mounted in both winter and summer and unpredictable fluctua- 1 2 on 2 replicate wood stakes with 9 buttons ranging from tions in summer rainfall. The rate of temperature change 2 3 ground level to 80-cm high. I-buttons were set to record at Kluane (2 °C per 50 years on average) is very high by 3 4 temperature at 4-hour intervals year-round. global standards of changing temperatures. 4 5 5 Statistical analyses were carried out in NCSS 10 Mammal community changes 6 (Number Crunching Statistical System, https://www. 6 7 ncss.com). Multiple regression models to investigate The overall change in the composition of the herbi- 7 8 hypotheses about the relative importance of weather, vore community at Kluane over the past 30 years is il- 8 9 predation and food factors were first investigated with lustrated in Fig. 2. We averaged the best estimates of 9 10 All Possible Regressions in NCSS 10 and then tested the population size of the 7 most common forest herbi- 10 11 with Robust Regressions, following the general recom- vore species, ignoring the species with low abundance 11 12 mendations of Mac Nally (2000). When several models in this part of the boreal zone. We converted these av- 12 13 13 seemed of equal value, we used AICc to determine the erage population estimates to energy use with the equa- 14 best model following the methods of Anderson (2008). tions provided in Nagy et al. (1999). If we use estimates 14 15 For all food-related statistical models we tested current of standing biomass, we obtain a similar picture (Boons- 15 16 annual production as well as that of the previous sum- tra et al. 2016), but we consider that energy use is a bet- 16 17 mer for possible predictive value because some berries ter estimate of ecosystem composition. There are 3 ma- 17 18 carry over from one year to the next. jor differences in the herbivore community over this 30- 18 19 year period. First, the contribution of small rodents to 19 20 20 RESULTS energy flow in the herbivore community increased from 21 9% in 1986–1996 to 38% in 2003–2014. This change 21 22 Climate change in the Kluane Region is the major topic of this paper. Second, arctic ground 22 23 squirrels [Urocitellus parryii (Richardson, 1825)] de- 23 24 Winter temperatures (November to April) have been clined 95%, changing status from very common to rare 24 25 increasing on average since 1970 by 0.47 °C per decade over this period (Werner et al. 2015). Third, snowshoe 25 26 (see Fig. 2 in Boonstra et al. 2018). Summer tempera- hares decreased 55% in energy flow in the decade from 26 27 tures during the growing season are increasing more 1986–1996 to 2003–2014, associated with lower aver- 27 28 slowly, at 0.26 ºC per decade (Fig. 1). Summer rainfall age densities after 2000 (Krebs et al. 2018a). Thus, the 28 29 (June to August) is highly variable and over the 1970 overall energy use of mice and voles has increased 73% 29 30 to 2018 period shows no significant change with a high in the period from 1986–1996 to 2003–2014. We ad- 30 31 coefficient of variation (CV = 40%). The net result is a dress the possible reasons for this change in mouse and 31 32 slowly changing climate with temperatures increasing vole populations in this paper. 32 33 33 34 34 35 35 36 36 37 37 38 38 39 39 40 40 41 41 42 42 43 43 44 44 45 45 46 Figure 1 Summer growing season aver- 46 47 age temperatures, 1970–2018, from Haines 47 48 Junction Weather Station. The overall re- 48 49 gression is: Summer temperature = −40.28 49 50 + (0.0261 × year), n = 49, R2 = 0.14, CV = 50 51 7.9%. Summer is June 1 to August 31. 51

© 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, 531 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd C. J. Krebs et al.

1 Changes in mice and vole population abundance only say that in terms of energy flow they are negligible. 1 2 The relative abundances of the most common species 2 3 Four rodent species in this boreal forest are rare- 3 ly seen or caught: Phenacomys intermedius Merriam, of small mammals has changed dramatically during the 4 last 45 years (Fig. 3). In the 1970s, the deer mouse Per- 4 5 1889, Zapus hudsonius (Zimmermann, 1780), Lemmus 5 sibiricus (Kerr, 1792) and Synaptomys borealis (Rich- omyscus and the red-back vole Myodes were the 2 domi- 6 nant species, each comprising approximately 40% of the 6 7 ardson, 1828). These species are nearly impossible to 7 study because they are so rare, and we have little infor- rodent numbers. By the 2000s dominance had switched 8 so that Myodes comprised approximately two-thirds of 8 9 mation on their ecology in this ecosystem. Their trappa- 9 bility may be extremely low relative to the main species the rodent numbers and Peromyscus only 12%. The 4 10 Microtus species, by contrast, have not changed very 10 11 but we know they are rare from both live-trapping and 11 snap-trapping studies (Krebs & Wingate 1976). We can much, being consistently as a group comprising approx- 12 imately 20% of the numbers of small mammals. 12 13 13 14 14 15 15 16 16 17 17 18 18 19 19 20 20 21 21 22 22 23 23 24 24 25 25 26 26 Figure 2 Energy flow averages for the 27 27 mammalian herbivores in the Kluane bore- 28 28 al forest ecosystem for the periods 1986– 29 29 1996 and 2003–2014. Both these periods 30 30 equally span a full 10-year snowshoe hare 31 cycle from the beginning of 1 cycle to the 31 32 beginning of the next. Energy flow esti- 32 33 mated from average density of each spe- 33 34 cies and the equations in Nagy et al. (1999). 34 35 35 36 36 37 37 38 38 39 39 40 40 41 41 42 42 43 43 44 44 Figure 3 Relative abundance of 10 species 45 in the small mammal community at Kluane 45 46 from the start of our studies (1973–1977) 46 47 to the more recent studies. The period 1978 47 48 to 1986 is not included because we did not 48 49 have adequate comparable data on the less 49 50 common species of small mammals for 50 51 that time period. 51

532 © 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Climate change impact on a boreal forest community

1 Red-back voles creased predator abundance coincided with an increase 1 2 rather than a decrease in Myodes abundance. Ermine 2 Figure 4 shows the changes in the abundance of the 3 abundance as indexed by snow tracking also increased 3 red-back vole Myodes from 1973 to 2018. There is a 4 after 2000 (Fig. 5). There was no association between 4 clear 3–4-year cycle in this species in the Kluane area 5 ermine abundance and marten abundance from 1986 to 5 (spectral density peaks at 3.67 years). There is a sugges- 6 2000 (because marten were virtually absent) but a high 6 tion of an increase in amplitude of peak densities over 7 correlation (r = 0.74) from 2000 to 2018. If predation is 7 time. A linear regression with peak year as the indepen- 8 a significant variable affecting rodent numbers, Myodes 8 dent variable and peak density as the dependent variable 9 density should have declined after 2000 but, instead, it 9 is suggestive but not statistically significant. A clearer 10 increased. This suggests that rodent numbers drive pred- 10 2 sample t-test with a division point at the year 2000 is 11 ator numbers but that the reverse effect is not as strong. 11 highly significant. From 1973 to 1999, 7 peaks averaged 12 12 9.7 Myodes per ha, while from 2000 to 2018, 5 peaks Winter snow conditions could also be related to Myo- 13 13 averaged 19.5 Myodes per ha. Significance values are des demography. However, neither maximum snow 14 14 meaningless for these comparisons since they were se- depth nor average levels of winter snow were closely 15 15 lected by inspection of Fig. 4, but the effect size is large, related to overwinter decline of Myodes (rs = −0.10 and 16 16 a doubling of peak Myodes density after 2000. 0.09, n = 45 years). More detailed local snow data from 17 2000 to 2018 shows the same result (Fig. S1), at best 17 Marten were nearly absent at Kluane before the year 18 a weak positive relationship between maximum snow 18 2000 (9). Their arrival coincided with the break point in 19 depth in winter and winter population changes in Myo- 19 20 peak abundance of Myodes, but the surprise was that in- 20 21 21 22 22 23 23 24 24 25 25 25 26 26 27 20 27 28 Figure 4 Spring and fall density estimates 28 perha. 29 for the red back vole Myodes rutilus at 15 29 30 Kluane, 1973 to 2018. Spring = green cir- 30 10 31 cles; fall = red squares. Peak years are Myodes 31 shaded in grey. Year tics are 1 January of

32 No. 32 33 each year. The horizontal dashed line is the 5 33 average density of Myodes over the entire 34 34 time period (6 per ha). Confidence inter- 0 35 35 vals are 95%. 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 36 36 37 37 38 38 39 39 40 100 40 41 41 42 80 42 43 43 Marten 44 60 Ermine 44 Figure 5 Snow track indices of small Red fox 45 mammal predators at Kluane, 1987 to 45 46 2018. Marten became abundant only af- 40 46 47 ter 2000, and at the same time ermine in- 47

48 creased greatly in abundance. Red fox- day per 100 km per tracks No. 20 48 49 es have never been common in the Kluane 49 50 boreal forest. Error bars are 95% confi- 0 50 51 dence limits. 87/8888/8989/9090/9191/9292/9393/9494/9595/9696/9797/9898/9999/0000/0101/0202/0303/0404/0505/0606/0707/0808/0909/1010/1111/1212/1313/1414/1515/1616/1717/18 51

© 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, 533 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd C. J. Krebs et al.

1 des. We have no evidence that snow depth in winter Deer mice 1 2 2 plays a strong role in determining the amount of popula- Deer mice (Peromyscus) now make up a small part 3 3 tion change of Myodes over the winter non-breeding pe- of the biomass of the small rodents at Kluane, approxi- 4 4 riod. mately 10%. Their population changes have been inex- 5 5 By measuring winter temperature at ground level in plicable for us (details in Krebs et al. 2018b; see Fig. S2 6 6 winter, we could ask if winter severity measured by dai- in Supplementary Materials). We have carried out a de- 7 7 ly temperatures below −10ºC independent of snow depth tailed statistical analysis of our deer mouse data. Sum- 8 8 could be related to overwinter disappearance of Myodes. 9 9 We began these measurements only in 2004 and with 14 10 10 years of data we found no relationship between winter 11 11 temperature severity at ground level and Myodes over- 12 12 winter disappearance (r = −0.09). Whatever the cause of 13 1 13 2 14 the variation in Myodes overwinter declines, it does not R = 0.67 14 appear to directly simply relate to the winter tempera- 15 0 15 16 ture under the snow at ground level. There was no sign 16 of ground level icing events in our area, although some 17 -1 17 18 mid-to-late winter rain-on-snow events did occur. 18

19 We combined all the measured variables into 1 mul- rate instantaneous predicted -2 19 20 tiple regression analysis to measure the relative strength decline overwinter of 20

21 of food and predator numbers on both summer growth -3 21 22 and winter declines of Myodes. For summer growth, the 22 23 major predictors were the berry crops of Empetrum ni- rutilus Myodes -3 -2 -1 0 1 23 24 grum, Vaccinium vitis-idaea, Geocaulon lividum and Myodes rutilus observed instantaneous rate 24 25 A. uva-ursi in the current year (Table 1). For winter de- of overwinter decline 25 26 clines, the major predictors were the E. nigrum and A. Figure 6 Prediction of red-back vole Myodes overwinter de- 26 27 rubra crops of the previous summer and the abundance cline in relation to observed winter decline. Prediction was 27 achieved by the equation given in Table 1 using crops of ber- 28 of ermine and marten in the current winter (Fig. 6, Table 28 ries the previous summer, 1997 to 2018, ermine and marten 29 1). Thus, both food supplies and predator numbers, but 29 30 abundance. Berries included Arctostaphylos rubra and Em- 30 not snow conditions or winter temperature, affect popu- petrum nigrum. Overwinter decline was calculated as the in- 31 lation changes in Myodes. 31 32 stantaneous rate of spring density in year t + 1/fall density in 32 year t. 33 33 34 34 35 35 36 36 37 Table 1 Stepwise predictions of Myodes rutilus summer population growth and winter population declines 37 38 Predictive Multiple regression Sample size R2 Variables considered but omitted 38 39 variable (years) 39 40 † 40 Summer Summer increase = 9.63 + 0.1074 22 0.58 May, June, July and August temperature, Arctostaphylos 41 population Vaccinium − 0.1078 Empetrum uva-ursi, Arctostaphylos rubra, Shepherdia canadensis, 41 42 increase – 0.2405 Geocaulon Mushroom crop, ermine and marten abundance in 42 43 previous winter 43 44 Winter Overwinter_change† = 21 0.67 Winter temperature, winter average snow depth, winter 44 45 population −0.670 + 0.0537 × Empetrum‡ maximum snow depth, previous summer (Arctostaphylos 45 46 decline − 0.1757 × A. rubra‡ − 0.0888 × uva-ursi, Shepherdia canadensis, mushroom crops) 46 47 Ermine + 0.0685 × Marten 47 48 48 All possible regressions were computed in NCSS 10 to determine the optimal number of variables for prediction using Mallow’s Cp 49 followed by stepwise regression to select the relevant variables, followed by robust multiple regression in NCSS 10. †Summer in- 49 ‡ 50 crease = fall density/spring density. Overwinter change = loge(spring density of t+1/fall density of year t). Berry production of pre- 50 51 vious summer. 51

534 © 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Climate change impact on a boreal forest community

1 mer data on the berry crops of A. uva-ursi and Geo- Microtus voles 1 2 caulon lividus along with snow tracks of ermine in the 2 The 4 Microtus species that are found at Kluane pres- 3 previous winter could predict summer increases in deer 3 ent yet another enigma to our understanding. Figure 7 4 mice. However, we could find no correlation with tem- 4 shows population trends in this species group over the 5 perature, snow levels, or the abundance of ermine or 5 past 33 years, and Table 3 gives the predictive regres- 6 marten that could predict the size of the winter decreas- 6 sions for this species group. The clear 3–4-year popula- 7 es over the 19 years for which we have detailed data 7 tion cycles are expected from these meadow voles, but 8 (Table 2). Since 1995, deer mice have never exceeded 8 the unusual pattern shown is that the dominant species 9 4 animals per ha, a density so low that predators might 9 at the peak changes from cycle to cycle, even though 10 not make a living hunting them and diseases might not 10 we are live trapping on exactly the same areas over this 11 easily transmit among them (see Luis et al. 2012). We 11 time period. The minor Microtus species do not disap- 12 have found no evidence of hantaviruses in Kluane Pero- 12 pear but in the peaks 95–100% of the captures are the 13 myscus and it is unlikely that these virus diseases could 13 dominant species (Fig. 7). Microtus oeconomus (Pal- 14 transmit in our populations with such a low density. 14 las, 1776) and M. pennsylvanicus (Ord, 1815) can be 15 15 16 16 17 17 Table 2 Stepwise predictions of Peromyscus maniculatus summer population growth and winter population declines 18 18 2 19 Predictive Multiple regression Sample size R Variables considered but omitted 19 20 variable (years) 20 21 Summer Summer_increase† = 19 0.74 May, June, July August temperature, Empetrum nigrum, 21 22 population 3.686 + 0.0706 × Arctostaphylos Arctostaphylos rubra, Shepherdia canadensis, Vaccinium 22 23 increase uva-ursi − 0.0982 × Geocaulon − vitis-idaea, mushroom crop, marten abundance in 23 24 0.0348 × Ermine previous winter 24 25 Winter No predator or food supply or 19 0.19 Ermine and marten abundance, winter temperature, 25 26 population temperature variables could predict winter average snow depth, winter maximum snow depth, 26 27 decline overwinter decline for Peromyscus previous summer (Empetrum nigrum, Arctostaphylos 27 rubra, Arctostaphylos uva-ursi, Shepherdia canadensis, 28 28 mushroom crops) 29 29 30 All possible regressions were computed in NCSS 10 to determine the optimal number of variables for prediction using Mallow’s Cp 30 followed by stepwise regression to select the relevant variables, followed by robust multiple regression in NCSS 10. †Summer in- 31 31 crease = fall density/spring density. Overwinter change = log (spring density of t + 1/fall density of year t). 32 e 32 33 33 34 12 34 35 miurus 35 36 oeconomus oeconomus 36 10 miurus pennsylvanicus 37 37 38 longicaudus 38 8 pennsylvanicus 39 39 40 40 6 pennsylvanicus 41 per hectare oeconomus 41 y pennsylvanicus miurus 42 oeconomus 42 4 oeconomus 43 pennsylvanicus 43 Densit 44 44 45 2 45 46 46 47 0 47 48 1986 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016 2019 48 49 Figure 7 Autumn population density of Microtus spp. in the boreal forest at Kluane, 1986 to 2018. A 3–4-year cycle is clear, but 49 50 the species present at the peak changes over time. Error bars are 95% confidence limits. (Updated from Fig. 6 in Krebs et al. 2018b 50 51 with permission.) 51

© 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, 535 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd C. J. Krebs et al.

1 difficult to tell apart in the field (Boonstra et al. 2001) nation fits this Microtus community. The challenge this 1 2 and the peak densities of 1993, 1998 and 2017 involved suggestion faces is that of doing experiments with the 2 3 a mixture of these 2 Microtus species. The peak abun- very low densities of these Microtus species (except for 3 4 dance of Microtus has also increased since 2000, like the peak years). 4 5 the pattern found in Myodes (Fig. 4). We come away not understanding the processes driv- 5 6 Why should this species replacement pattern occur? ing these changes: in the end we have a group of Micro- 6 7 We can find no one who has studied voles who has de- tus species who appear to have little direct interactions 7 8 scribed such a pattern of cyclic species replacement. through interference competition yet persist with differ- 8 9 There are 2 possible models of how this pattern could ent habitat requirements and possible food differences. 9 10 arise. First, there may be a competitive dominance hi- 10 Changes in predator populations 11 erarchy among these species. We do not know if there 11 12 is direct interference competition between any of these The major predators of all these small rodents are er- 12 13 species, except for our studies on M. miurus (Osgood, mine and marten, as well as the great horned owls [Bubo 13 14 1901) and M. oeconomus, in which we could demon- virginianus (Gmelin, 1788)], the boreal owl [Aegoli- 14 15 strate no competitive release when we removed one of us funereus (Linnaeus, 1758)], the northern hawk owl 15 16 them and observed the response of the other (Galin- [Surnia ulula (Linnaeus, 1758)], and a variety of other 16 17 do & Krebs 1985a,b). We could see no clear patterns of raptors (Doyle & Smith 2001). For the 1986–1996 peri- 17 18 species interference in this small rodent community. If od, we have very good quantitative data on bird preda- 18 19 there was a size dominance in the Microtus guild, M. tors to correlate with the changes in vole and mice num- 19 20 longicaudus (Merriam, 1888) (average adult weight 50 bers (Doyle & Smith 2001; Rohner et al. 2001), and we 20 21 g) should win. The average weight of the other Micro- were able to continue these observations only for great- 21 22 tus species ranges from 25 to 32 g. A second model to horned owls until 2010. The overall mammalian preda- 22 23 explain Fig. 7 could be random species replacement in tor energy flow period has changed dramatically during 23 24 a patchy network. All these Microtus prefer grassland, the past 40 years. Marten accounted for 0.1% of the en- 24 25 which occurs in small patches throughout the Kluane ergy flow in the 1986–1996 period and then increased to 25 26 boreal forest. If populations go locally extinct in the low account for approximately 7% of the energy flow from 26 27 phase of these 3–4-year cycles, the next colonizer could 2003 to 2014. Over this same time period, the contribu- 27 28 become the winner for the next peak. Such metacommu- tion of ermine to overall predator energy flow tripled. 28 29 nity dynamics (Holyoak et al. 2005) implicates an array Figure 8 shows the overall average contribution to ener- 29 30 of possible mechanisms from dispersal dynamics to ran- gy flow in the predator populations for which we have 30 31 dom patch extinction among interacting species within good data. There is a suggestion that over this 30-year 31 32 a community. Much more experimental work and land- time period predation rates might have increased on 32 33 scape analyses will be needed to determine if this expla- small rodent populations. 33 34 34 35 35 36 36 37 Table 3 Stepwise predictions of Microtus spp. summer population growth and winter population declines. Berry production of the 37 38 previous summer was used for each berry species as a possible predictive variable for winter population changes 38 39 Predictive Multiple regression Sample size R2 Variables considered but omitted 39 40 variable (years) 40 41 Summer Summer_increase† = 7.593 21 0.59 May, June, July, August temperature, Empetrum 41 42 population + 1.5532 × Previous spring density nigrum, Arctostaphylos uva ursi, Arctostaphylos 42 43 increase − 0.3148 × Geocaulon − 0.1604 × rubra, Shepherdia canadensis, Vaccinium vitis-idaea, 43 44 Ermine + 0.2781 × Marten mushroom crop 44 45 Winter Winter_decline = −0.5562 19 0.19 Marten abundance, winter temperature, winter average 45 46 population + 0.02743 × A. uva-ursi snow depth, winter maximum snow depth, previous 46 47 decline + 0.02169 × Shepherdia − 0.1053 × summer (Empetrum nigrum, Arctostaphylos rubra, 47 48 Geocaulon − 0.02820 × Ermine Vaccinium vitis-idaea, mushroom crops) 48

49 All possible regressions were computed in NCSS 10 to determine the optimal number of variables for prediction using Mallow’s Cp 49 50 followed by stepwise regression to select the relevant variables, followed by robust multiple regression in NCSS 10. †Summer in- 50

51 crease = fall density/spring density. Overwinter change = loge(spring density of t + 1/fall density of year t). 51

536 © 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Climate change impact on a boreal forest community

1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 11 12 12 Figure 8 Energy flow averages for the 13 13 mammalian carnivores in the Kluane bore- 14 al forest ecosystem for the periods 1986– 14 15 1996 and 2003–2014. Both these periods 15 16 contain a full 10-year snowshoe hare cycle. 16 17 Energy flow estimated from the estimat- 17 18 ed average density of each species and the 18 19 equations in Nagy et al. (1999). The small 19 20 band between marten and wolf is the ener- 20 21 gy flow to ermine. 21 22 22 23 23 24 24 25 25 3. The population fluctuations of the 3 dominant ro- 26 DISCUSSION 26 dent species are not disappearing as a result of climate 27 27 Given these data, we can answer the 5 questions change. Both Myodes and Microtus spp. are increasing 28 28 posed in the Introduction. in abundance as time progresses, in correlation with in- 29 29 1. Populations of M. rutilus and Microtus spp. fluctu- creased primary production. 30 ate periodically in 3–4-year cycles, but P. maniculatus 30 31 4. Predation and food supply can both be implicat- 31 does not. M. rutilus has been challenged in some short- ed as partial causes of population change in Myodes 32 term studies as a non-cyclic vole in the boreal zone of 32 33 and Microtus and are associated with summer popula- 33 North America (e.g. Boonstra & Krebs 2006, 2012), tion changes in Peromyscus. The puzzle is that as small 34 but this conclusion does not fit our long-term Yukon 34 35 mammal predators have increased, so has rodent popu- 35 data, which is clearly cyclic. The closely-related Myo- lation abundance, the opposite prediction that one would 36 des glareolus in Fennoscandia is clearly cyclic in that 36 37 expect from predation theory if predation was a major 37 region, and we would predict that M. rutilus is cyclic factor causing population changes. The missing factor in 38 across the boreal forest zone of North America. P. ma- 38 39 the equation is social behavior, which can cause repro- 39 niculatus has never been observed to be cyclic so this ductive suppression (Gilbert et al. 1986) as well as pre- 40 conclusion is not new. 40 41 vent or increase infanticide (Millar 2007; Eccard et al. 41 42 2. Populations of the 3 major species at Kluane are 2011). These results are consistent with the experimen- 42 43 only weakly correlated. Autumn population densities of tal findings of Johnsenet al. (2017). 43 44 Myodes and Microtus are significantly correlated (r = 5. Boonstra and Krebs (2006) suggested that M. ru- 44 45 0.77) but neither of these is correlated with Peromyscus tilus population changes were linked to snowshoe hare 45 46 fall numbers (r = 0.26 to 0.35). We suspect this correla- cycles via nutrient cycling. Our data are inconsistent 46 47 tion of Myodes and Microtus is partly driven by ermine with this suggestion. With 42 years of data, we have 47 48 and marten predation, as well as possible raptor preda- found no significant correlation of snowshoe hare densi- 48 49 tion for which we do not have good data. A similar cor- ties with current Myodes, Microtus or Peromyscus den- 49 50 relation has been noted in Fennoscandia (Hansson & sities, nor are any of these rodent densities correlat- 50 51 Henttonen 1988). ed with hare densities lagged by one or two years. Any 51

© 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, 537 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd C. J. Krebs et al.

1 short-term correlation between snowshoe hare numbers such, our study could be a boreal forest control in which 1 2 and those of small rodents we would suspect to be spu- the climate drivers and the human drivers are not con- 2 3 rious. founded. 3 4 There is a general feeling among conservation biolo- The largest possible changes we can see in the next 4 5 gists that biodiversity is being lost on an ever-increasing 100 years that would greatly disrupt this part of the bo- 5 6 scale and that most populations that have been studied real zone are emergency events such as extensive forest 6 7 are declining in abundance (Estes et al. 2011; McRae fires or changes in land use that compromise the integri- 7 8 et al. 2017; Ripple et al. 2017; but see Vellend 2017). ty of intact forests (e.g. widespread logging or road net- 8 9 Added to this view is the overall observation that cli- works). Unless disruptions are very extensive, we can 9 10 mate change is occurring and is most severe in the arc- see no problems in ecosystem recovery, with the cave- 10 11 tic regions (IPCC 2013). We could expect from these 2 at that the time scale of recovery may be well over 100 11 12 views that we should see ecosystem disruptions in the years. 12 13 13 Kluane boreal forest. Our observations suggest the con- We do not wish to project our positive conclusions 14 14 trary view. Populations of small rodents in this ecosys- for small rodents onto the larger species in this eco- 15 15 tem are increasing in overall abundance and we have no system. The harvest of Canada lynx, grizzly and black 16 16 information on species losses among these mammals. bears, moose, caribou, and thin-horn sheep are co-man- 17 17 Vegetation (in particular shrubs) is growing more and in aged by First Nations and other governments with a 18 18 spite of a bark-beetle attack on white spruce trees in the conservation objective. Given the problems that large 19 19 1990s trees continue to grow and the forest is regener- mammals face in most parts of the Earth, our area is 20 20 ating (Berg et al. 2006; Krebs et al. 2014). Dirzo et al. again a contrast at present (Ripple et al. 2014; Boonstra 21 21 (2014) noted that in many ecosystems if large mammals et al. 2018) because none of these large mammal spe- 22 22 were eliminated or greatly reduced, rodent populations cies are under threat. 23 23 increased. Our data do not fit this paradigm exactly be- The bottom line of our analysis is very positive for 24 cause rodents have increased but large mammals con- 24 25 the continuation of this small mammal community in 25 tinue to be present at moderate numbers in this Yukon the Kluane region. We know that to be correct because 26 community. 26 27 of our long-term monitoring and we stress again that in 27 If our data are an anomaly within the global pattern 28 all ecosystems the “devil is in the details” and these de- 28 of trophic downgrading, several reasons can be given 29 tails can only be known by hard work and continued 29 in explanation. Northern terrestrial ecosystems are par- 30 monitoring. 30 ticularly slow moving because of the severe winters, 31 31 and, despite rapid climate change, the growing season 32 CONCLUSION 32 is limited to approximately 12 weeks from late May to 33 33 mid-August. The species in the boreal ecosystem are all Our 46-year study of these small rodents is a mod- 34 34 post-glacial colonists from approximately 12 000 years el because it is long-term with a consistent methodolo- 35 35 ago, when the glaciers receded and recolonization be- gy on fixed sites undisturbed except by climate change. 36 36 gan. As such, they are likely to be nearer the general- We have rigorously documented profound changes that 37 37 ist end of the spectrum than specialist species. Perhaps would never have been observed without this long-term 38 38 the greatest contrast with more temperate and tropical effort. It has both given us deep insights and deep puz- 39 39 conservation issues is that terrestrial ecosystems in the zles. It has also documented that the signature of climate 40 40 Kluane region have had minimal human impact. Resi- change is clear in these pristine forests. Despite all our 41 41 dent First Nation peoples have carried out subsistence efforts, there are large unknowns in this ecosystem. The 42 42 harvesting for millennia, and while the influx of set- birds of prey have not been studied in the continuous 43 43 tlers over the past 120 years has varied with econom- detail we would have liked. Great-horned owls are sug- 44 44 ic booms, total numbers have stayed below 1000. There gested to be declining but we do not know the causes. 45 45 is no commercial forestry but some firewood gathering, All the migratory birds are particularly difficult to clas- 46 46 no agriculture industry, no domestic cattle, and a small sify with respect to needed conservation actions and at 47 47 mining footprint. The analysis by Murray et al. (2017) present we have no data on their abundances. The plant 48 48 suggests that for the next 100 years the Kluane ecosys- base of the Kluane region is well known but few experi- 49 49 tem portion of the boreal forest will be the least affect- mental studies have been done, and less long-term mon- 50 50 ed by climate-induced forest fragmentation and loss. As itoring than is desirable. In consequence our plea is for 51 51

538 © 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Climate change impact on a boreal forest community

1 ecosystem monitoring to follow in the footsteps of cli- were essential to this long-term research program, and 1 2 mate monitoring with adequate resources so that in fu- we thank Andy and Carole Williams, Lance Goodwin 2 3 ture we can construct an analog of the Intergovernmen- and Sian Williams for their assistance. 3 4 tal Panel on Climate Change (IPCC) for the ecosystems 4 5 of the Earth. REFERENCES 5 6 Once there is adequate and continuing funding, there 6 7 are still 2 unresolved problems in recommending more Anderson DR (2008). Model Based Inference in the Life 7 8 long-term monitoring studies. One is where to do them, Sciences: A Primer on Evidence. Springer, New York. 8 9 and the other is how long they should be carried for- Berg EE, David Henry J, Fastie CL, De Volder AD, 9 10 ward. The where question is most easily answered be- Matsuoka SM (2006). Spruce beetle outbreaks on the 10 11 cause you need a protected area to carry out long-term Kenai Peninsula, Alaska, and Kluane National Park 11 12 monitoring, although that is not always easy to provide. and Reserve, Yukon Territory: Relationship to sum- 12 13 The how long question is more difficult, and the ques- mer temperatures and regional differences in dis- 13 14 tion is whether you are learning new concepts and in- turbance regimes. Forest Ecology and Management 14 15 sights from the continuing work. If you adopt the weath- 227, 219–32. 15 16 er station paradigm, you should never specify an end Boonstra R, Krebs CJ (2006). Population limitation of 16 17 date because monitoring is an ongoing requirement for the northern red-backed vole in the boreal forests 17 18 environmental management. of northern Canada. Journal of Animal Ecology 75, 18 19 19 Our Kluane monitoring program can be criticized 1269–84. 20 20 for ignoring the rare rodent species that could be the Boonstra R, Krebs CJ (2012). Population dynamics of 21 21 concern of conservation ecology. From a conservation red-backed voles (Myodes) in North America. Oeco- 22 22 viewpoint, these rare species are an important part of logia 168, 601–20. 23 23 biodiversity. From an ecosystem viewpoint, rare species 24 Boonstra R, Krebs CJ, Gilbert S, Schweiger S (2001). 24 are typically of little consequence to community dy- 25 10 voles and mice. In: Krebs CJ, Boutin S, Boons- 25 namics. If you cannot do everything within your budget, 26 tra R, eds. Ecosystem Dynamics of the Boreal Forest: 26 we think it is more important to study the dominant spe- 27 The Kluane Project. Oxford University Press, New 27 cies because they are the key to ecosystem change. For 28 York, pp. 215–39. 28 this particular boreal forest ecosystem, the 2 missing el- 29 Boonstra R, Andreassen HP, Boutin S et al. (2016). 29 ements that need study are the fungi with their mycor- 30 Why do the boreal forest ecosystems of Northwest- 30 rhizal associates and the impacts of climate change on 31 ern Europe differ from those of Western North Amer- 31 the shrubs and ground vegetation that provide the food 32 ica? BioScience 66, 722–34. 32 base for all of the herbivores. From a regional perspec- 33 33 tive, the linkage of the alpine tundra ecosystem with Boonstra R, Boutin S, Jung TS, Krebs CJ, Taylor S 34 34 the boreal forest ecosystem is an almost unstudied ele- (2018). Impact of rewilding, species introductions 35 35 ment with consequences that can flow in both altitudi- and climate change on the structure and function of 36 36 nal directions as the climate shifts. There is much left to the Yukon boreal forest ecosystem. Integrative Zool- 37 37 do and northern Canada provides some of the few eco- ogy 13, 123–38. 38 38 systems on Earth that are ruled by climate change rather Cornulier T, Yoccoz NG, Bretagnolle V et al. (2013). 39 39 than by human disturbances. Europe-wide dampening of population cycles in key- 40 stone herbivores. Science 340, 63–6. 40 41 41 ACKNOWLEDGMENTS Dirzo R, Young HS, Galetti M, Ceballos G, Isaac NJB 42 (2014). Defaunation in the Anthropocene. Science 42 43 We thank Mark O’Donoghue, Elizabeth Hofer, Pe- 345, 401–6. 43 44 44 ter Upton and many students who have worked over the Doyle FI, Smith JNM (2001). 16 raptors and scaven- 45 years to capture these measurements. We thank 2 re- 45 46 gers. In: Krebs CJ, Boutin S, Boonstra R, eds. Eco- 46 viewers for critical comments that improved the man- system Dynamics of the Boreal Forest: The Klu- 47 uscript. Research funding was provided by the Natural 47 48 ane Project. Oxford University Press, New York, pp. 48 Science and Engineering Research Council of Canada 377–404. 49 (RB, CK and SB). The facilities of the Kluane Lake Re- 49 Eccard JA, Jokinen I, Ylönen H (2011). Loss of den- 50 search Station of the Arctic Institute of North America 50 51 sity-dependence and incomplete control by domi- 51

© 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, 539 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd C. J. Krebs et al.

1 nant breeders in a territorial species with density out- Krebs CJ (2013). Population Fluctuations in Rodents. 1 2 breaks. BMC Ecology 11, 16. University of Chicago Press, Chicago. 2 3 Elton C (1942). Voles, Mice and Lemmings: Problems in Krebs CJ, Wingate I (1976). Small mammal communi- 3 4 Population Dynamics. Clarendon Press, Oxford. ties of the Kluane Region, Yukon Territory. Canadian 4 5 5 Efford MG, Fewster RM (2013). Estimating population Field-Naturalist 90, 379–89. 6 6 size by spatially explicit capture–recapture. Oikos Krebs CJ, Wingate I (1985). Population fluctuations in 7 7 122, 918–28. the small mammals of the Kluane Region, Yukon 8 8 Territory. Canadian Field-Naturalist 99, 51–61. 9 Estes JA, Terborgh J, Brashares JS et al. (2011). Trophic 9 10 downgrading of Planet Earth. Science 333, 301–6. Krebs CJ, Boutin S, Boonstra R (2001). Ecosystem Dy- 10 11 Fuller WA (1969). Changes in numbers of three species namics of the Boreal Forest: The Kluane Project. Ox- 11 12 of small rodent near Great Slave Lake, N.W.T., Can- ford University Press, New York. 12 13 ada, 1964–1967, and their significance for general Krebs CJ, Boonstra R, Cowcill K, Kenney AJ (2009). 13 14 population theory. Annales Zoologici Fennici 6, 113– Climatic determinants of berry crops in the boreal 14 15 44. forest of the southwestern Yukon. Botany 87, 401–8. 15 16 Fuller WA (1977). Demography of a subarctic popula- Krebs CJ, Boonstra R, Boutin S et al. (2014). Trophic 16 17 tion of Clethrionomys gapperi: Numbers and surviv- dynamics of the boreal forests of the Kluane Region. 17 18 al. Canadian Journal of Zoology 55, 42–51. Arctic 67, 71–81. 18 19 Galindo C, Krebs CJ (1985a). Habitat use by singing Krebs CJ, Boonstra R, Boutin S (2018a). Using exper- 19 20 voles and tundra voles in the southern Yukon. Oeco- imentation to understand the 10-year snowshoe hare 20 21 logia 66, 430–6. cycle in the boreal forest of North America. Journal 21 22 Galindo C, Krebs CJ (1985b). Habitat use and abun- of Animal Ecology 87, 87–100. 22 23 dance of deer mice: Interactions with meadow voles Krebs CJ, Boonstra R, Kenney AJ, Gilbert BS (2018b). 23 24 and red-backed voles. Canadian Journal of Zoology Hares and small rodent cycles: a 45-year perspective 24 25 63, 1870–9. on predator-prey dynamics in the Yukon boreal for- 25 26 26 Gilbert BS, Krebs CJ, Talarico D, Cichowski DB (1986). est. Australian Zoologist 39, 724–32. 27 27 Do Clethrionomys rutilus females suppress matura- Likens GE (1989). Long-term Studies in Ecology: Ap- 28 28 tion of juvenile females? Journal of Animal Ecology proaches and Alternatives. Springer Verlag, New 29 29 55, 543–52. York. 30 30 31 Hansson L, Henttonen H (1988). Rodent dynamics as Luis AD, Douglass RJ, Hudson PJ, Mills JN, Bjørnstad 31 32 community processes. Trends in Ecology and Evolu- ON (2012). Sin Nombre hantavirus decreases surviv- 32 33 tion 3, 195–200. al of male deer mice. Oecologia 169, 431–9. 33 34 Holyoak M, Leibold MA, Holt RD (2005). Metacom- Mac Nally R (2000). Regression and model-building in 34 35 munities: Spatial Dynamics and Ecological Commu- conservation biology, biogeography and ecology: the 35 36 nities. University of Chicago Press, Chicago. distinction between – and reconciliation of – ‘predic- 36 37 Hughes BB, Beas-Luna R, Barner AK et al. (2017). tive’ and ‘explanatory’ models. Biodiversity & Con- 37 38 Long-term studies contribute disproportionately to servation 9, 65–71. 38 39 ecology and policy. Bioscience 67, 271–81. McRae L, Deinet S, Freeman R (2017). The diversi- 39 40 IPCC 2013 (2014). IPCC Fifth Assessment Report: Cli- ty-weighted living planet index: controlling for tax- 40 41 mate Change 2013: The Physical Science Basis. Con- onomic bias in a global biodiversity indicator. PLoS 41 42 tribution of Working Group I to the Fifth Assessment ONE 12, e0169156. 42 43 Report of the Intergovernmental Panel on Climate Millar JS (2007). Nest mortality in small mammals. 43 44 Change. Cambridge University Press, Cambridge, Ecoscience 14, 286–91. 44 45 45 UK. [Cited 1 Jun 2019.] Available from URL: https:// Murray DL, Peers MJL, Majchrzak YN et al. (2017). 46 46 doi.org/10.1017/CBO9781107415324 Continental divide: Predicting climate-mediated frag- 47 47 Johnsen K, Boonstra R, Boutin S, Devineau O, Krebs mentation and biodiversity loss in the boreal forest. 48 48 CJ, Andreassen HP (2017). Surviving winter: Food, PLoS ONE 12, e0176706. 49 49 but not habitat structure, prevents crashes in cyclic 50 Nagy KA, Girard IA, Brown TK (1999). Energetics of 50 vole populations. Ecology and Evolution 7, 115–24. 51 free-ranging mammals, reptiles, and birds. Annual 51

540 © 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd Climate change impact on a boreal forest community

1 Review of Nutrition 19, 247–77. ane Region. Arctic 67 (Suppl. 1), 98–107. 1 2 O’Donoghue M, Boutin S, Hofer EJ, Boonstra R (2001). Vellend M (2017). The Biodiversity Conservation Para- 2 3 14 other mammalian predators. In: Krebs CJ, Boutin dox. American Scientist 105, 94–101. 3 4 4 S, Boonstra R, eds. Ecosystem Dynamics of the Bo- Werner JR, Krebs CJ, Donker SA, Boonstra R, Sher- 5 5 real Forest: The Kluane Project. Oxford University iff MJ (2015). Arctic ground squirrel population col- 6 6 Press, New York, pp. 324–36. lapse in the boreal forests of the Southern Yukon. 7 7 Radchuk V, Ims RA, Andreassen HP (2016). From in- Wildlife Research 42, 176–84. 8 8 dividuals to population cycles: The role of extrinsic 9 9 and intrinsic factors in rodent populations. Ecology 10 SUPPLEMENTAL MATERIALS 10 97, 720–32. 11 Additional supporting information may be found in 11 12 Ripple WJ, Estes JA, Beschta RL et al. (2014). Status 12 and ecological effects of the world’s largest carni- the online version of this article at the publisher’s web- 13 site. 13 14 vores. Science 343, 1241484. 14 Figure S1 Red-backed voles (Myodes rutilus) in- 15 Ripple WJ, Wolf C, Newsome TM et al. (2017). World 15 stantaneous rate of change from autumn to the follow- 16 scientists’ warning to humanity: A second notice. 16 ing spring in relation to winter maximum snow depth, 17 BioScience 67, 1026–8. 17 2000–2018 at Kluane Lake. There is an indication of a 18 Rohner C, Doyle FI, Smith JNM (2001). 15 great 2 18 weak positive relationship (R = 0.15, P = 0.10. Average 19 horned owls. In: Krebs CJ, Boutin S, Boonstra R, 19 snow depth over winter is highly correlated with maxi- 20 eds. Ecosystem Dynamics of the Boreal Forest: The 20 mum snow depth (r = 0.82, n = 18). 21 Kluane Project. Oxford University Press, New York, 21 22 pp. 339–76. Figure S2 Deer mouse (Peromyscus maniculatus) 22 autumn population density 1976–2018 at Kluane Lake, 23 Slough BG (1989). Movements and habitat use by trans- 23 with 95% confidence limits. No deer mice were live 24 planted marten in the Yukon Territory. Journal of 24 trapped in the forest between 1990 and 1995 in spite of 25 Wildlife Management 53, 991–7. 25 26 extensive trapping. These data are discussed in Krebs 26 Turkington R, McLaren JR, Dale MRT (2014). Herba- et al. (2018b). Figure updated from Krebs et al. (2018b) 27 ceous community structure and function in the Klu- 27 28 with permission. 28 29 29 30 30 31 Cite this article as: 31 32 32 33 Krebs CJ, Boonstra R, Gilbert BS, Kenney AJ, Boutin S (2019). Impact of climate change on the small mammal 33 34 community of the Yukon boreal forest. Integrative Zoology 14, 528–41. 34 35 35 36 36 37 37 38 38 39 39 40 40 41 41 42 42 43 43 44 44 45 45 46 46 47 47 48 48 49 49 50 50 51 51

© 2019 The Authors. Integrative Zoology published by International Society of Zoological Sciences, 541 Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd