Explaining bank vole cycles in southern Norway from bilberry reports and climate

Transcription

1 Oecologia (2006) 147: DOI /s POPULATION ECOLOGY Vidar Sela s Explaining bank vole cycles in southern Norway from bilberry reports and climate Received: 14 June 2005 / Accepted: 23 November 2005 / Published online: 13 December 2005 Ó Springer-Verlag 2005 Abstract Correlations between mast fruiting of bilberry Vaccinium myrtillus and peak levels of Clethrionomysvoles have been reported from both Norway and Finland, but there has been a discussion whether this is a bottom-up or a top-down relationship. In a multiple regression model, 65% of the variation in a bilberry production index calculated from game reports from southern Norway could be explained by the berry index of the two preceding years and climate factors acting during key stages of the flowering cycle. High vole populations in previous years did not contribute to explain the fluctuation in berry production. I used the selected model and climate data to predict bilberry production for the period Predicted berry indices of the current and previous year explained 38% and the total amount of precipitation in May June explained 16% of the variation in a log-transformed snaptrapping index of bank vole Clethrionomys glareolus The vole index was not related to any of the climate variables used to predict berry production. This pattern supports the hypothesis that vole cycles are generated by changes in plant chemistry due to climatesynchronized mast fruiting. Keywords Clethrionomys Æ Food quality Æ Predation Æ Vaccinium Introduction Although the most recent research on 3 4 year population cycles of voles and lemmings have focused on Communicated by Hannu Ylonen V. Selås Department of Ecology and Natural Resource Management, Norwegian University of Life Sciences, P.O. Box 5003, 1432 A s, Norway vidar.selas@umb.no Tel.: Fax: trophic interactions, there is still no consensus with regard to the ultimate cause for these cycles. The impact of predation has been addressed in several field studies, confirming that predators influence rodent population levels (e.g. Korpima ki and Norrdahl 1998; Ekerholm et al. 2004), but sufficient strong support to conclude that predators are needed to generate rodent cycles is still lacking. The predation hypothesis was actually contradicted in a field experiment on field vole Microtus agrestis in England (Graham and Lambin 2002; see also Oli 2003; Lambin and Graham 2003), but supporters of the hypothesis have so far questioned the generality of this study (Korpima ki et al. 2003). An alternative to the predation hypothesis is that 3 4 year rodent cycles are initiated by mast-induced changes in plant chemistry (White 1993; Sela s 1997), and only modified by predation. According to White (1993), low availability of amino acids is usually a limiting factor for growth and reproduction of herbivores. Like other plant stress responses that represent a metabolic cost, the production of a high seed crop will require increased mobilization and transport of soluble amino acids, which is the easiest available N-source for herbivores (White 1993). Mast fruiting may also require resources that would otherwise have been used for chemical defence. Even slight changes in levels of free amino acids or compounds that inhibit the absorption of these nutrients may have a significant effect on herbivore reproduction and survival if the availability of amino acids is usually below a critical threshold for herbivores protein budget (White 1993). One obvious problem in studies of rodent cycles is the difficulty in manipulating these large-scale phenomena experimentally. Another problem, in my view, is that the complexity of plant chemistry and our limited understanding of the role of nutritional and defensive plant compounds on herbivore performance make experimental studies on plant herbivore interactions difficult. An alternative approach is to examine whether the assumptions for leading hypotheses are in accordance with empirical data. Even though correlative studies can

2 626 only be indicative with regard to underlying processes, the patterns they reveal cannot be ignored in our search for the explanation of rodent cycles. After having presented significant correlations between indices of seed crops of bilberry Vaccinium myrtillus and autumn population levels of bank vole Clethrionomys glareolus from southern Norway (Selås 1997), similar to that found for bilberry and grey-sided vole C. rufocanus in Finland (Laine and Henttonen 1983), the discussion has been whether bilberry plants are generating vole cycles (e.g. Sela s et al. 2002a), or whether voles, regulated by predators, are generating cycles in berry crops (e.g. Oksanen et al. 1999). According to the latter view, berry production fluctuates in synchrony with vole grazing intensity because repairing damaged shoot systems has priority over reproduction (Tolvanen et al. 1993). However, as periods of vole cycles often varies from 3 to 5 years, it should be possible to reveal who depends on who from time series analyses. The problem seems to be that we lack sufficient long-term bilberry vole series to give convincing results. During , information on bilberry production in Norway was given in annual game reports (Myrberget 1982; Sela s 1997, 2000a). Unfortunately, there are few records on bank vole densities from this period. However, assuming that (1) there is an endogenous flowering cycle in bilberry plants which is not or only slightly influenced by vole grazing, (2) the seed production of individual bilberry plants are synchronized through weather events acting during key stages of this cycle and (3) voles are influenced by berry production through plant chemistry, it should be possible to use information on bilberry production from 1932 to 1977 and climate data to forecast vole cycles after If, on contrary, fluctuations in bilberry production depend mainly on vole grazing, former berry production should be irrelevant for vole cycles after Here I use bilberry information from game reports to predict bilberry and vole fluctuations in the recent 25 years from an area where annual population estimates on bank vole have been obtained from 1980 onwards. Materials and methods The study area is the eastern part of Aust-Agder County in southern Norway, which is situated in the boreo-nemoral zone. The area covers approximately 3,000 km 2, and is dominated by a fine-grained mosaic of young, medium-aged and old deciduous, mixed and coniferous forest stands (Sela s 1997, 2000a; Sela s et al. 2002a). During , the forest age classes <40, and >80 years constituted approximately 30, 20 and 50% of the forest area, respectively (Tomter 1994; Tomter et al. 2001). Snap-trapping conducted with 400 1,500 trap-nights each autumn (with at least 50% of the traps in old forest) from 1980 onwards has revealed a significant 3 4 year fluctuation pattern in bank vole populations in this area (Sela s 1997; Sela s et al. 2001, 2002a; Framstad 2003). In game reports from 1932 to 1977, the bilberry production was designated as lower-than-medium (value 0), medium (value 1) or higher-than-medium (value 2). I supplemented game reports with similar evaluations given in local newspapers by representatives of Aust-Agder Berry Company, the Norwegian Agricultural Price Reporting Office and local agricultural or forestry authorities, given that these persons were not identical to those who had signed game reports that year. The total number of annual reports (game reports and other) ranged from 2 (1944, 1974) to 30 (mean 9.33, SD=6.65). For each year, I calculated a bilberry production index as the mean value from all reports. In years with many bilberries, a large proportion of the reporters should be expected to evaluate the bilberry production to be higher than average, giving a bilberry index close to two, while in years with low production most reporters should evaluate the production to be lower than average, giving a bilberry index close to zero. With this bilberry index as response variable I constructed a multiple linear regression model with previous berry production, the presence of high vole populations in previous years and climate variables acting during key stages of the flowering cycle as possible predictor variables. The criterion for model selection was that the difference in AIC value between the selected model and the model with the lowest AIC value was less than 2.0 (Burnham and Anderson 1998). Game reports were not used to calculate vole indices, because they usually did not distinguish between mice and voles. In southern Norway, wood mouse Apodemus sylvaticus populations peak after years with high production of acorns, whereas this food resource seems to be less important for bank voles (Sela s et al. 2002a). Hence, in addition to reports on rodent damages in forestry and agriculture, which are always caused by voles, information on acorn production (see Sela s 2000b) made it possible to distinguish between the vole and mouse peak years. According to the top-down hypothesis for bilberry vole cycles, a high berry production in year t should require that there were low vole populations in year t 1 and t 2. Hence, I used the presence of high vole populations in year t 1 ort 2, as well as the presence in each of these years and in year t 3, as dummy variables (value 0 or 1) in the bilberry model. Climate data were provided by the Norwegian Meteorological Institute. For the entire study period, data were available as monthly means (temperatures, snow depth), monthly total (precipitation) or monthly maximum (snow depth). Precipitation and snow depth were measured within the study area, while a sufficient long data series on temperatures had to be taken from a station located outside the study area, approximately 50 km west of the snap-trapping areas (see Sela s 2000a). The sunspot series used were the yearly means calculated

3 627 as the average of daily means ( DATA/yearssn.dat). Figure 1 shows expected effects of climate and previous reproduction on berry production in year t. Although resources may be accumulated for several years before masting (Isagi et al. 1997), I simplified the model by assuming that summer temperatures (mean June September), sunspot numbers and berry production in year t 2 were the main factors determining the amount of accumulated resources important for flower bud induction in year t 1. The rational for using sunspot numbers was that increased levels of surface UV-B radiation in periods with low sunspot activity may represent a cost because of increased production of UV-Bprotective phenolics (Sela s et al. 2004). I further assumed that berry production in year t 1 negatively affects flower bud induction. I used mean August temperatures as an index for weather conditions during flower bud induction, which is reduced at high temperatures (Spann et al. 2004). During winter, bilberry shoots may suffer from dehardening, induced by high temperatures in periods with thin snow cover (O gren 1996). To account for this negative effect on flower buds, I used the mean temperature and maximum snow depth from January, and the interaction effect of these two variables. Because snow reduces the negative effect of high temperatures, the interaction effect was expected to be positive. Furthermore the risk of flower damage due to frost in May (Belonogova 1988; Kuchko 1988) may be modified by snow, because much snow delays flowering (Tolvanen 1997). For the period , there was a significant positive relationship between maximum snow depth in April in the study area and the date of the first recorded bilberry flowering (May) in the neighbouring county Vest-Agder (R 2 =0.19, P=0.009, n=44). Hence, I used both maximum snow depth in April and mean temperatures in May as explanatory variables. Finally, drought during June or July, which has been common in the study area, will impede berry developing (Kuchko 1988). However, too much rain may also be negative (Phoenix et al. 2001). I therefore calculated a hydrothermal ratio index (total precipitation divided by Fig. 1 Expected effects (+ or ) of climate and previous reproduction on seed crops of bilberry in year t mean temperature; Myllyma ki et al. 1985) for the months June and July separately, and then used only the lowest of these indices in the analysis. For the period , the only significantly correlated climate variables were the two snow depth indices (r=0.55, P<0.001). From the selected bilberry regression model , I predicted annual bilberry indices for the period To verify the reliability of the predicted bilberry indices, I also searched for statements on bilberry production in local newspapers for the period Bilberry indices and predicted berry indices , as well as the climate variables from these two periods, were tested for periodicity in spectral density analyses. The bank vole snap-trapping index was log-transformed before analyses to improve normality of residuals. Then I compared the vole index with the predicted bilberry index and all climate variables used in the bilberry model, both the current and previous year. Because favourable weather condition in spring and early summer is likely to hasten vole reproduction and thus increase autumn numbers, regardless of the phase of the population cycle, I also tested for relationships with mean temperatures and the total amount of precipitation in May June. These variables were, however, significantly correlated (r = 0.43, P=0.033, n=25). In the multiple regression analysis, the criteria for variable selection were as for the bilberry analysis. Results There were no significant periodicity in the climate series from , but there was a tendency for a 3.5-year cycle in the calculated bilberry index (Fisher s Kappa=4.98, P=0.099, n=46; Fig. 2). In the bilberry index there was a significant negative autocorrelation at a time lag of 2 years (r = 0.30, P=0.046, n=44). The bilberry index was also negatively correlated with the summarized bilberry index of the two preceding years (r = 0.34, P=0.026, n=44). There was a significant positive relationship with maximum snow depth in April (R 2 =0.10, P=0.037, n=44) and almost so with the lowest hydrothermal ratio index in June July (R 2 =0.08, P=0.057, n=44; Fig. 2). There were no relationships between the bilberry index and the presence of high vole populations in previous years (t 1 ort 2: R 2 =0.03, P=0.214; t 1: R 2 <0.01, P=0.744; t 2: R 2 =0.02, P=0.315; t 3: R 2 =0.03, P=0.251). The best regression model included the summarized bilberry index of the two previous years and all climate variables except summer temperatures year t 2 and May temperatures year t, although there was an almost significant positive effect of summer temperatures (Table 1). The selected bilberry model for , which explained 65% of the variation in the bilberry index, was used to predict annual bilberry indices for the period Values below 0 (1989 and 1992) and above 2

4 628 Fig. 2 Indices of population levels of bank vole and bilberry production in Aust-Agder, southern Norway. The bank vole indices are the number of animals trapped per 100 trap-nights in autumn. The bilberry indices (thick line) are calculated from game reports (see text), whereas the bilberry indices (thin line) are predicted from the bilberry regression model (Table 1). The figure also shows fluctuations in the two climate variables that were best related with bilberry production ; maximum snow depth in April and the lowest of the hydrothermal ratio (total precipitation divided by mean temperature) in June and July (1980) were substituted with 0 and 2, respectively. The predicted bilberry index showed a significant 3.4-year periodicity (Fisher s Kappa = 6.82, P=0.002, n=27). The index peaked in , 1984, 1987, 1990, 1993, 1997 and 2001, and was very low in 1983, 1989 and 1992 (Fig. 2). The result for 1980, 1983, 1984, 1989 and 1992 was in accordance with newspaper interviews with local forestry or agricultural authorities. Newspapers also reported high berry production in 1987, 1990, 1993 and 1997, but information sources were not given. Only in 1981, there was obviously a deviation between predicted (high) and real berry production, which was low according to local forestry authorities. There were no 3 4 years cycles in climate variables from 1978 to 2004, but there was a significant 7-year periodicity in mean August temperatures (Fisher s Kappa=5.01, P=0.038, n=27). The log-transformed bank vole index from 1980 to 2004 showed a 3.5-year periodicity (Fisher s Kappa=5.08, P=0.023, n=25). The only significant autocorrelation was at a time lag of 2 years (r = 0.51, P=0.009). Bank vole populations peaked in autumn 1981, 1985, 1988, 1991, 1994 and , and there also was a minor peak in (Fig. 2). The log-transformed vole index could not be explained by any of the climate variables obtained in the bilberry model, either in the current or the previous year. The results were the same whether or not the vole indices of the two preceding years were included in the analyses. The vole index was, however, negatively related to the total amount of precipitation in May June (R 2 =0.16, P=0.045, n=25). There also was a significant positive relationship with the predicted bilberry index of the current (R 2 =0.18, P=0.033, n=25) and previous year (R 2 =0.16, P=0.049, n=25). In a multiple regression model, both the bilberry variables and precipitation in Table 1 Linear multiple regression model with a berry production index for bilberry Vaccinium myrtillus (year t) in Aust-Agder, southern Norway, calculated from game reports (n=44), as response variable Explanatory variable Regression coefficints SE df R 2 AIC P Intercept Berries, year t 1 + year t <0.001 Max snow depth, April (cm) Hydrothermal ratio, June or July <0.001 Temperatures August, year t Temperatures January Max snow depth, January Temp snow, January Sunspots, year t Summer temperatures, year t Temperatures May Presence of voles year t 1 ort Presence of voles year t Presence of voles year t Presence of voles year t Possible explanatory variables are the summarized berry production index of the two previous years, climatic factors acting during key stages of the flowering cycle and the presence of high vole populations in previous years (see text). Regression coefficients with standard errors refer to the chosen model. The selected variables are listed in the direction they were obtained in a forward selection procedure, and cumulative values are given for R 2 and AIC. For each of the other variables, the values refer to the model where also this variable was selected

5 629 May June contributed significantly to explain the bank vole index (Table 2). If the presumably too high bilberry index of 1981 was reduced from 1.9 to , the explanatory power of the model is increased from 54 to 61%. Discussion Except from summer temperatures 2 years earlier and May temperatures, all climate variables contributed significantly to explain the bilberry index from 1934 to 1977, given that also previous reproduction was obtained in the model. That berry production and sunspots numbers appeared to be more important than summer temperatures for the estimated resource storage prior to flower bud induction could perhaps be expected, since these two variables varied much more than summer temperatures. The lack of a significant relationship between berry production and mean temperatures in May indicates that this climate index failed to reveal the absence or presence of frost during the flowering period, which in the study area varies from early May to early June. The presence of high vole populations in previous years could not explain the bilberry index. But even though possible negative effects of vole herbivory were not adjusted for, the selected regression model explained 65% of the variation in the bilberry index from 1932 to Furthermore the fact that the predicted indices for fitted well with available information on bilberry production from this second period indicates that the model captured the most important factors for bilberry production in southern Norway. Moreover for masting trees, annual flowering or seed crops have been successfully predicted by use of climate data and previous reproduction (Sela s et al. 2002b; Ranta et al. 2005). The predicted bilberry indices of the current and previous year explained 18 and 20%, respectively, and May June precipitation 16% of the variation in the logtransformed bank vole index from 1980 to The almost equal contribution of bilberries in the current and previous year in the regression model indicates that vole population growth was high both in the berry peak year and 1 year later. Due to the exponential nature of population growth, this will generate a marked peak in vole populations 1 year after a bilberry peak, as seen in the untransformed time series. Also spring precipitation explained much of the variation in the vole index, indicating that weather variables may act as disturbing factors in regular rodent fluctuations. It could be argued that the regression of vole numbers against bilberry production is only an indirect method to check whether climate affect voles. However, it should be noted that except from snow cover, the explanatory variables used in the bilberry model are not likely to have any direct effects on bank vole performance, and that a necessary condition for obtaining significant relationships between berries and voles was that previous berry production was included in the bilberry regression model. Indeed, bank vole populations could not be explained by any of the climate factors used to predict bilberry production. The patterns reported here are difficult to interpret as a result of a top-down relationship, i.e. that the bilberry vole cycle is caused by predation or vole grazing. Korpima ki et al. (2003) questioned the generality of the field experiment of Graham and Lambin (2002) because of some fundamental differences between vole cycles in England and northern Europe. The differences addressed were cycle amplitudes, spatial synchrony and synchrony between different small mammal species. The question then is whether southern Norway differs from the study areas of Korpima ki and co-workers in these respects. As elsewhere in Fennoscandia, vole populations in southern Norway have been times higher in peak than in low vole years during Also the geographical synchrony has been comparable to that of other parts of Fennoscandia (Christiansen 1983). And as in the study areas of Korpima ki et al. (2003), there has usually been a synchronous population fluctuation of Clethrionomys and Microtus voles (Christiansen 1983). Furthermore the similar fluctuation pattern of bilberry and Clethrionomys-voles in Norway and Finland (Laine and Henttonen 1983) suggests that there is no rational for expecting any differences with regard to the ultimate cause for vole cycles in different parts of northern Europe. Correlative studies cannot reveal underlying mechanisms, for instance whether bank voles respond to nutrients in berries, to increased amino acid transport or to reallocation of secondary compounds. But regardless of the ultimate cause for a bilberry-dependent vole cycle, Table 2 Linear multiple regression model with the log-transformed snap-trapping index of bank vole Clethrionomys glareolus (year t) in Aust-Agder, southern Norway, (n=25), as response variable Explanatory variable Regression coefficients SE df R 2 AIC P Intercept Predicted berries, year t Predicted berries, year t Rain May June (mm) The explanatory variables are total amount of precipitation in May June and bilberry production indices predicted from the bilberry regression model from 1932 to 1977 (Table 1). The variables are listed in the direction they were obtained in a forward selection procedure, and cumulative values are given for R 2 and AIC

6 630 there probably are two necessary conditions for such cycles to be generated and continued. First, the environment must be too harsh for annual seed cropping. In mild climates, rowan Sorbus aucuparia will fruit each year, whereas in Fennoscandia, the species has a highly variable and synchronous seed production (Kobro et al. 2003). If a similar pattern exists for bilberry, highamplitude cycles of Clethrionomys-voles should be expected only at higher altitudes and latitudes. Second, a sufficient high frequency of years with unfavourable conditions for flowering or seed ripening may be required in order to synchronise individual plants. If climate remains stable for a long time period, individual plants could still have a cyclic seed production, but increasing individual asynchrony may eventually result in stable vole populations. Climate changes may alter these two conditions and thus account for some recently observed deviations in population fluctuations of small rodents in Fennoscandia (e.g. Strann et al. 2002; Ho rnfeldt 2004). There are no doubts that predation influences vole population levels, and that vole grazing influences the performance of plants, but the present study suggests that neither predation nor vole grazing is needed to generate vole cycles. In a recent paper on the effect of predator removal on vole dynamics, Ekerholm et al. (2004) conclude that predation may not be a necessary condition for microtine cycles. The main role of predators seems to be to depress vole populations in the decreasing phase of the population cycle, so that the amplitude of the vole cycle is enhanced. If the underlying mechanisms for vole cycles are unknown, such an effect could mislead an observer to believe that predators are generating the cycle. Voles may affect the cycle period if they have a sufficient negative effect on bilberry plants to prolong the time needed to prepare for a new seed crop, and may, thus underserved, be accused for creating cycles in bilberry flowering. Interestingly, food plant longevity appears to be a better predictor for the period length in herbivore cycles than herbivore size or taxonomy (Ho gstedt et al. 2005). There have been some recent attempts to apply a general specialist/generalist predator hypothesis both to the 3 4 year rodent cycles and the 10-year cycles of snowshoe hare Lepus americanus and forest moths feeding on woody plants (Klemola et al. 2002; Korpima ki et al. 2004). However, as populations of mountain hare Lepus timidus have peaked at 10-year intervals in the absence of specialist predators (Hewson 1976; Sela s 2006), fluctuating food quality may be the ultimate cause also for 10-year herbivore cycles. The mast depression hypothesis seems to be less appropriate for 10-year cycles of moths feeding on birch Betula pubescens (Klemola et al. 2003), but another stress factor that may alter plant chemistry and thus affect herbivore performance is fluctuating levels of UV-B radiation during the sunspot cycle (Sela s et al. 2004; Sela s 2006). Anyway, to achieve further progress in the field of cyclic herbivore populations, we should focus not only on animal ecology, but also on plant physiology and ecology. Acknowledgements I am grateful to Erik Framstad and Tor K. Spidsø for providing data from their rodent snap-trapping studies. The idea for the analyses presented here emerged from comments, questions and suggestions made by different referees on my former papers on mast cropping and herbivore cycles. These usually anonymous referees deserve some credit for forcing me to improve my analyses and for inspiring me to continue my studies on herbivore cycles. References Belonogova TV (1988) Yield forecasting, optimization of berry harvesting in the forests of Southern Karelia, USSR. Acta Bot Fenn 136:19 21 Burnham KP, Anderson DR (1998) Model selection and multimodel inference. A practical information theoretic approach. Springer, Berlin Heidelberg New York, 488 pp Christiansen E (1983) Fluctuations in some small rodent populations in Norway Holarct Ecol 6:24 31 Ekerholm P, Oksanen L, Oksanen T, Schneider M (2004) The impact of short-term predator removal on vole dynamics in an arctic-alpine landscape. Oikos 106: Framstad E (2003) Monitoring programme for terrestrial ecosystems. Ground vegetation, epiphytes, small rodents and birds in the monitoring sites, 2002 (in Norwegian, with English abstract). NINA Oppdragsmelding 793, 62 pp Graham IM, Lambin X (2002) The impact of weasel predation on cyclic field-vole survival: the specialist predator hypothesis contradicted. J Anim Ecol 71: Hewson R (1976) A population study of mountain hares (Lepus timidus) in north-east Scotland from J Anim Ecol 45: Ho gstedt G, Seldal T, Breistøl A (2005) Period length in cyclic animal populations. Ecology 86: Ho rnfeldt B (2004) Long-term decline in numbers of cyclic voles in boreal Sweden: analysis and presentation of hypotheses. Oikos 107: Isagi Y, Sugimura K, Sumida A, Ito H (1997) How does masting happen and synchronize? J Theor Biol 187: Klemola T, Tanhuanpäa M, Korpima ki E, Ruohoma ki K (2002) Specialist and generalist natural enemies as an explanation for geographical gradients in population cycles of northern herbivores. Oikos 99:83 94 Klemola T, Hanhima ki S, Ruohoma ki K, Senn J, Tanhuanpää M, Kaitaniemi P, Ranta H, Haukioja E (2003) Performance of the cyclic autumnal moth, Epirrita autumnata, in relation to birch mast seeding. Oecologia 135: Kobro S, Søreide L, Djønne E, Rafoss T, Jaastad G, Witzgall P (2003) Masting of rowan Sorbus aucuparia L. and consequences for the apple fruit moth Argyresthia conjugella Zeller. Popul Ecol 45:25 30 Korpima ki E, Norrdahl K (1998) Experimental reduction of predators reverses the crash phase of small-rodent cycles. Ecology 79: Korpima ki E, Klemola T, Norrdahl K, Oksanen L, Oksanen T, Banks P, Batzli GO, Henttonen H (2003) Vole cycles and predation. Trends Ecol Evol 18: Korpima ki E, Brown PR, Jacob J, Pech RP (2004) The puzzles of population cycles and outbreaks of small mammals solved? Bioscience 54: Kuchko AA (1988) Bilberry and cowberry yields and the factors controlling them in the forests of Karelia, USSR. Acta Bot Fenn 136:23 25 Laine K, Henttonen H (1983) The role of plant production in microtine cycles in northern Fennoscandia. Oikos 40: Lambin X, Graham IM (2003) Testing the specialist predator hypothesis for vole cycles. Trends Ecol Evol 18:493

7 631 Myllyma ki A, Hansson L, Christiansen E (1985) Models for forecasting population trends in two species of microtine rodent, Microtus agrestis and Clethrionomys glareolus. Acta Zool Fenn 173: Myrberget S (1982) Production of some wild berries in Norway (in Swedish, with English summary). Fauna Flora 77: O gren E (1996) Premature dehardening in Vaccinium myrtillus during a mild winter: a cause for winter dieback? Funct Ecol 10: Oksanen T, Schneider M, Rammul U, Hambäck P, Aunapuu M (1999) Population fluctuations of voles in North Fennoscandian tundra: contrasting dynamics in adjacent areas with different habitat composition. Oikos 86: Oli MK (2003) Population cycles of small rodents are caused by specialist predators: or are they? Trends Ecol Evol 18: Phoenix GK, Gwynn-Jones D, Callaghan TV, Sleep D, Lee JA (2001) Effects of global change on a sub-arctic heath: effects of enhanced UV-B radiation and increased summer precipitation. J Ecol 89: Ranta H, Oksanen A, Hokkanen T, Bondestam K, Heino S (2005) Masting by Betula-species; applying the resource budget model to north European data sets. Int J Biometeorol 49: Selås V (1997) Cyclic population fluctuations of herbivores as an effect of cyclic seed cropping of plants: the mast depression hypothesis. Oikos 80: Selås V (2000a) Seed production of a masting dwarf shrub, Vaccinium myrtillus, in relation to previous reproduction and weather. Can J Bot 78: Selås V (2000b) Is there a higher risk for herbivore outbreaks after cold mast years? An analysis of two plant/herbivore series from southern Norway. Ecography 23: Selås V (2006) UV-B-induced plant stress as a possible cause for 10- year hare cycles. Popul Ecol (in press) Selås V, Sonerud GA, Histøl T, Hjeljord O (2001) Synchrony in short-term fluctuations of moose calf body mass and bank vole population density supports the mast depression hypothesis. Oikos 92: Selås V, Framstad E, Spidsø TK (2002a) Effects of seed masting of bilberry, oak and spruce on sympatric populations of bank vole (Clethrionomys glareolus) and wood mouse (Apodemus sylvaticus) in southern Norway. J Zool Lond 258: Selås V, Piovesan G, Adams JM, Bernabei M (2002b) Climatic factors controlling reproduction and growth of Norway spruce in Southern Norway. Can J For Res 32: Selås V, Hogstad O, Kobro S, Rafoss T (2004) Can sunspot activity and ultraviolet-b radiation explain cyclic outbreaks of forest moth pest species? Proc R Soc Lond B 271: Spann TM, Williamson JG, Darnell RL (2004) Photoperiod and temperature effects on growth and carbohydrate storage in southern highbush blueberry interspecific hybrid. J Am Hortic Sci 129: Strann K-B, Yoccoz NG, Ims RA (2002) Is the heart of Fennoscandian rodent cycle still beating? A 14-year study of small mammals and Tengmalm s owls in northern Norway. Ecography 25:81 87 Tolvanen A (1997) Recovery of the bilberry (Vaccinium myrtillus L.) from artificial spring and summer frost. Plant Ecol 130:35 39 Tolvanen A, Laine K, Pakonen T, Saari E, Havas P (1993) Aboveground growth response of the bilberry (Vaccinium myrtillus L.) to simulated herbivory. Flora 188: Tomter SM (1994) Statistics of forest conditions and resources in Norway (in Norwegian, with English summary). Report, Norwegian Institute of Land Inventory, A s Tomter SM, Eriksen R, Aalde H (2001) Statistics of forest conditions and resources in Aust-Agder (in Norwegian, with English abstract). The National Forest Inventory Report, Norwegian Institute of Land Inventory, A s White TCR (1993) The inadequate environment Nitrogen and the abundance of animals. Springer, Berlin Heidelberg New York