Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
O. Tenow, Department of Entomology, Swedish University of Agricultural Sciences, PO Box 7044, S-750 07, Sweden. E-mail: firstname.lastname@example.org
1In 1990–2003, during a complete 10-year outbreak cycle, the synchrony of the birch defoliating outbreaks of the geometrids Epirrita autumnata and Operophtera brumata was studied quantitatively in the northern part of the Fennoscandian mountain chain (the Scandes). Data were supplemented with similar data from 1964 to 1966 and historical information. A 30-year series of field data from one locality in southern Scandes made possible interregional comparisons.
2In 1991, outbreaks started in north-eastern Fennoscandia and moved westward like a wave and reached the outer coast of north-western Norway in about 2000. This wave is a new observation. In the same years, a previously documented outbreak wave moved southward along the Scandes.
3Outbreak periods have usually occurred around the middle of each decade. Seemingly unrelated population peaks at the decadal shift 2000 were reported from islands at the coast of north-western Norway. They are shown here to have been the final ripples of the east–west wave.
4At some localities, O. brumata peaked 2 years after E. autumnata. A lag of 1 or 2 years also occurred at the locality in southern Scandes. This interspecific time lag is a new observation. In accordance with the north–south wave, a time-lag of 1–2 years occurred between the fluctuations of northern and southern E. autumnata and O. brumata populations.
5The population peak of E. autumnata occurred 1 year earlier at one locality than at a nearby locality. This pattern and particular altitudinal shifts of the O. brumata population density at these localities repeated in two outbreak periods. This indicates that, for example, local climate may modify outbreak synchrony between nearby localities.
6At the same localities, O. brumata peaked first at one altitude and 1 or 2 years later at another altitude. This vertical lag is a new observation.
7E. autumnata shows fluctuation traits similar to some other cyclic animals, e.g. the larch budmoth in the European Alps, some European tetraonid birds and the Canadian snow-shoe hare. These similarities (and dissimilarities) in intra- and interspecific synchronies and causes of E. autumnata and O. brumata synchronies, regionally, locally and among the two species are discussed.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
About every 9–10th year, vast areas of birch (Betula pubescens Ehrh.) forests in Fennoscandia have been defoliated by caterpillars of Epirrita autumnata and/or Operophtera spp., particularly in the Fennoscandia mountain chain (the Scandes). Primarily, old forests are attacked (Bylund 1997; Ruohomäki et al. 1997). Periodic (‘cyclic’) outbreaks such as these have been recorded for several folivorous forest insects (for a review, see Myers 1988) and some game birds and mammals (Keith 1963). The 10-year cyclic outbreaks of the larch budmoth, (Zeiraphera diniana Gn), which defoliate subalpine larch (Larix decidua Mill.) in the European Alps (Baltensweiler & Fischlin 1988), are well known. Among mammals, the 10-year cyclic snowshoe hare (Lepus americanus Erxl.) is another example (Keith 1963).
The E. autumnata fluctuations share the characteristics described above: (i) outbreaks have been 10-year-cyclic in northern Fennoscandia (Neuvonen 1988); (ii) during some periods, the local outbreaks occurred concurrently all over the Scandes; (iii) during other outbreaks they were of the ‘moving type’ (Tenow 1972), i.e. over years they ‘spread’ like a wave from northern Norway and Finland along the Scandes to southern Norway or the reverse; and (iv) outbreaks have been synchronized with those of the sympatric O. brumata.
Qualitative information for 1862–1968 revealed 12 outbreak periods (I–XII) on a Fennoscandian scale (Tenow 1972). This periodicity has continued during the 1970s, 1980s and 1990s (Nilssen, Tenow & Bylund 2007). Consequently, these later periods have been numbered XIII, XIV and XV.
During most of the time, outbreaks have centred on the middle of each decade. However, in the 1970s, 1980s and 1990s, geometrid populations in the coastal area of north-westernmost Scandes have peaked at decadal shifts (around 1979, 1989 and 2000), i.e. maximally out-of-phase with outbreaks otherwise in northern Fennoscandia (Nilssen et al. 2007). These findings are based on qualitative information. In their study in 1999–2002, Ims et al. (2004) demonstrated such asynchrony quantitatively. Recently, Klemola et al. (2006) reported on spatial synchrony of E. autumnata population peaks across Fennoscandia.
In order to investigate whether or not outbreaks of E. autumnata and O. brumata are synchronous, in this paper we present quantitative and qualitative population data from northern Fennoscandia for 1990–2003. The synchrony of density fluctuations of both species are surveyed between localities and within locality. A short-term study in the north during the 1960s and the long-term study (1972–2001) in southern Scandes by Hogstad (2005) enabled us to conduct between-periods and between-regions comparisons, and thus to evaluate and discuss the degree of synchronization at different temporal and spatial scales.
Materials and methods
Caterpillars of E. autumnata and O. brumata pass five instars (I–V) during their growth. The E. autumnata caterpillar feeds openly on leaves, whereas the O. brumata caterpillar often lives in leaves spun together. When fully grown in July, caterpillars of both species leave the trees and pupate in the litter. For E. autumnata, the pupal stage lasts until August–September; in some areas for O. brumata it also lasts into October, when the moths emerge. Both species mate and lay eggs in autumn. Both sexes of E. autumnata and the male of O. brumata can fly, whereas the female has stunted wings and must therefore climb the tree to lay eggs (Haukioja et al. 1988; Peterson & Nilssen 1998). The eggs are placed singly on branches and twigs of the birch, exposed to the extremes of weather. The eggs overwinter and hatch at budburst. The eggs of E. autumnata freeze at −36 °C (their supercooling point, SCP) in midwinter (Nilssen & Tenow 1990). The corresponding SCP for O. brumata eggs is −35 °C (McPhee 1967). Both species occur over most of Fennoscandia but their outbreak distributions differ: O. brumata's in meadow birch forests along the western side of the Scandes and E. autumnata's mainly in heath birch forests along the eastern side, except for a gap in the cold Finnmarksvidda highland plain of north-eastern Norway (Tenow 1972). There, even the more cold-hardy E. autumnata eggs often freeze (Tenow & Nilssen 1990).
One southern cofluctuating geometrid, Agriopis aurantiaria Hb., occurs at warm localities in the north and was included in the study. It has a life cycle similar to that of E. autumnata and O. brumata (South 1961).
Reports on outbreaks and quantitative data on population density peaks were used to recognize the period(s) of outbreaks. The criteria of an ‘outbreak’ were severe damage to foliage and/or mass-occurrence of caterpillars. ‘Peaks’ are culminations in population densities either as outbreaks or as peaks below outbreak level, where caterpillar numbers (in a few cases moth numbers) were estimated quantitatively.
Historical documents on outbreaks in the 1990s and the early 2000s were available for Norway from http://www.nisk.no/skogskade/report. Information for Sweden was collected through personal communications with biologists and earth scientists, tourist guides and national wildlife guards visiting the outbreak area. Observations by the authors and information in newspapers provided complementary data. Outbreak and population peak years have been surveyed thoroughly by Finnish ecologists working in the outbreak area. Information for Finland has therefore been taken exclusively from published Finnish studies (for details, see Nilssen et al. 2007). The complete historical documentation is available for consultation in the Supplementary material (see http://www.blackwell-synergy.com).
field data sampling
Synchrony in E. autumnata and O. brumata fluctuations was evaluated from data on larval abundances sampled in a 30-year series from one locality (Budal) in southern Scandes (Hogstad 2005) and in decadal series from six localities in northern Scandes (Table 1) selected along an east–west transect from the Finnmarksvidda to the Ofoten–Vesterålen–Lofoten islands, the extreme western offset of northern Scandes (Fig. 1). The northern investigation is complemented by data from the 1960s, when a similar study was made (Tenow 1990).
Table 1. Localities and altitudes for quantitative samplings of Epirrita autumnata, Operophtera brumata and Agriopis aurantiaria caterpillars in the Scandes in 1964–66 and 1990–2003
2 determined by a pocket altimeter (Lambrecht 1304).
300, 360, 460
260, 360, 450
135, 200, 260, 350
50, 120, 230, 310, 390
120, 230, 310, 390
30, 110, 250
20, 120, 235
Larval population densities were estimated in June and/or July at Budal. Caterpillars were collected randomly along a horizontal line, 2–3 km in length. Caterpillars were swept with a net from branches within the lowest 4 m of birch trees (i.e. nearly the whole tree height). Each year five to 15 collections, each of 100 sweeps, were taken when the caterpillars had reached instars IV and V. The annual index of caterpillars is the mean number per 100 sweeps (Hogstad 1997). At each locality in northern Scandes, sampling stations were established at different altitudes up-slope. The number of such altitudinal stations was adjusted to the vertical extent of the locality (Table 1). In the first years six trees were sampled; six and 10 trees were sampled later at high and low population levels, respectively, at each altitudinal station. The trees were selected within a 30 × 30-m plot (later 50 × 50 m) by using the sides of the plot as a coordinating system and choosing the birch tree nearest to each of six or 10 randomized points. From each tree, one branch in mid-crown was clipped down to a sheet. Branches (on average with > 100 leaf-carrying shoots) were each put into a separate plastic bag. Shoots and caterpillars of each species were counted per branch and the larval number per branch expressed per 100 shoots. The annual index is the mean number of caterpillars per 100 shoots (n = 6 or 10) from each altitudinal station.
During the outbreak years 1964–66 (with some complements for 1963), density changes of the E. autumnata and O. brumata populations were monitored at two localities (Engmo, Gratangen; Fig. 1) at the same altitudinal stations as in the 1990s. Six or 10 trees were selected in a similar manner (for details, see Tenow 1990).
Each locality in northern Scandes was sampled within 1–2 days. Because the altitude difference between the highest and lowest station in the vertical profile could be > 200 m with a due phenological lag up-slope, the developmental stage of caterpillars could vary from instars I–III at the highest station to instars IV–V at the lowest station of a locality. A parallel lag in mortality causes an over-estimation of caterpillar densities at higher relative to lower altitudes (cf. Miller & Epstein 1986). In the 1960s this was overcome by revisiting localities after completed larval feeding to look for horizontal brown zones of defoliated forests. Such observations were also made in 1963 prior to the start of quantitative studies (Tenow 1972: 92).
mapping of outbreaks
Locations of all E. autumnata and O. brumata population peaks (for criteria, see ‘Historical documentation’) recorded qualitatively or quantitatively in northern Fennoscandia were mapped and dated by the year of occurrence (Fig. 3). Approximate isochrones were drawn to circumscribe peak locations of the same year. Such locations thus fall into the same ‘isochronal belt’ delimited by two successive isochrones. The closer to the coast, the narrower the belts appear. At nearby locations, the timing of peaks varied from no difference in time (central northern Finland) to 2–3 years (at the coast). This variation probably reflected a systematic trait rather than uncertainty in the timing of outbreaks or peaks.
In order to analyse population density changes vertically in slopes, population density data of the population peak years at each locality were log(X + 1)-transformed to meet assumptions of normality and homogeneous variance, and analysed in a repeated-measurement analysis of variance (anova) in proc mixed (sas version 9·1·3). The model density = year altitude, year × altitude was used. If the interaction term year × altitude was significant, the estimates of the solution for fixed effect of the interaction term was tested for pairs of altitudes and successive years in t-tests. This was conducted to determine whether the significant interaction effect was due to densities peaking at different altitudes in successive years. For some within-year comparisons, proc glm (sas) analyses followed by Tukey's multiple comparison tests were performed.
synchrony within region
Figure 2a shows the growth and decline of the larval populations in the northern Scandes during the outbreak period of the 1990s. The fluctuation changes are given as the mean density for all altitudinal stations of each locality per year. At Máze, the easternmost locality, only E. autumnata caterpillars were found; at Frihetsli, Engmo, Gratangen, Løbergsbukta and Melbu both E. autumnata and O. brumata occurred. At the two last (westernmost) localities, A. aurantiaria was added, indicating an increased diversity of the geometrid guild towards west.
Although the population densities varied strongly between localities (Fig. 2a), populations (E. autumnata) appear to have culminated first at the easternmost locality, Máze, in 1993 and later (E. autumnata and O. brumata) at the two westernmost localities, Løbergsbukta and Melbu in 1998. At the three intermediate localities, Frihetsli, Engmo and Gratangen, the populations of both species culminated between those two years, i.e. in 1994–96; however, O. brumata densities declined more slowly after the peak at the two last localities. The A. aurantiaria population at Løbergsbukta (and Melbu) peaked in about 2000. The time-lag in E. autumnata population peaks between the outermost localities to the east and the west was 5 years, half the time of an average population cycle. Population peaks were below outbreak numbers at Máze, Frihetsli, Løbergsbukta and Melbu but not at Engmo and Gratangen. The documentation of defoliated birch forests shows that the low peaks were contemporaneous with nearby outbreaks. On the Finnmarksvidda, severe defoliations by E. autumnata occurred in 1993 and 1994, 35 km south-west and 16 km north of Máze, respectively. In 1994, a defoliating outbreak of E. autumnata occurred 16 km south of Frihetsli. To the west, on the island of Hinnøy in Vesterålen, an O. brumata outbreak occurred in 1998, 5 km south of Løbergsbukta.
In addition there were severe, more distant, defoliations. The first recorded E. autumnata outbreak of the period was reported from north-eastern Fennoscandia in 1991–92 (Fig. 3). From this region, outbreaks seem to have ‘spread’ in several directions. E. autumnata outbreaks occurred in central northern Finland in 1992–94, at the north-eastern edge of the Finnmaksvidda and in northernmost Sweden in 1993–95. To the north, outbreaks of O. brumata occurred along the coast of Norway in 1994–95. Finally, in the far west, there were O. brumata population peaks along the Norwegian coast north of the Ofoten–Vesterålen–Lofoten area in about 1998–2000 (Ims et al. 2004). Hence, qualitative and quantitative data support our field data, showing that the population peaks along the studied transect were part of a wave of outbreaks moving from east to west in 1991–2000, i.e. with a time-lag of 9 years over the entire east–west distance. Outbreaks at the northern coast (latitude 71° N) in 1999 showed that waves also reached the far north with a considerable delay. In general, outbreak ‘spread’ may have slowed down in coastal areas, as indicated by the narrowing of isochronal belts toward the coast (Fig. 3).
synchrony within and among localities
The degree of synchrony in intraspecific population density fluctuations was related to the geographical distance between the localities compared. When the rate of change in population density of E. autumnata or O. brumata was correlated pairwise, the strongest associations (Pearson's r-values) between time-series were indicated for localities at a short distance (Fig. 4). The associations seem to vanish rapidly at distances of between 60 and 100 km. However, a negative trend can be discerned, indicated by significant correlations between the r-values for pairs of localities and the distance between them for both species (E. autumnata: Pearson's r = −0·744, P = 0·002, n = 14 and O. brumata: Pearson's r = −0·973, P = 0·000, n = 10).
The temporal relationships between E. autumnata, O. brumata and A. aurantiaria density fluctuations are illustrated in Fig. 2a. At Máze only E. autumnata occurred. At Frihetsli, Engmo and Gratangen, O. brumata lagged 1 or 2 years behind E. autumnata. At Løbergsbukta and Melbu such lags were lacking. Instead, A. aurantiaria lagged 1–2 years behind E. autumnata and O. brumata, demonstrating the possibility of a fluctuation lag among geometrids at the extreme west as well. Also at Budal, O. brumata lagged 1 year behind E. autumnata (Fig. 2b).
Larval densities did not always culminate concurrently at all altitudes in a slope. This is seen when population data for Gratangen and Engmo, monitored in the 1960s and 1990s, are tested in mixed-model anovas (Fig. 5). In addition to mainly significant effects of year and altitude in all series tested, the interaction term year × altitude was significant for O. brumata. For this species at Gratangen, the model's interaction term was significant for 1964–66 (d.f. = 5, 112, F = 3·99, P = 0·0023). This effect was explained mainly by the significant interactions when the slopes of density changes at altitudes of 50 m vs. 120 m and 50 vs. 230 were compared between 1965 and 1966 (t-value = −2·87 and t = −3·51, respectively, d.f. = 112 and P < 0·001 for both). This means that the peak densities at 120 m and 230 m occurred in 1965 followed by a decline, while densities increased at 50 m in 1966. At Engmo, a significant interaction was found over the years 1964–66 (d.f. = 4, 65, F = 6·92, P = 0·0001). The larval density increased from 1964 to a peak in 1965 at altitude 350 m (t-test of estimates 1964 and 1965: 350 m vs. 260 m d.f. = 65, t = 3·43 and 350 m vs. 200 m, P = 0·0009, d.f. = 65, t = 3·48, P = 0·0009), while densities were not changing at altitudes 260 m and 200 m. In 1966 densities had declined at all altitudes. In 1963 an O. brumata outbreak caused a browning of the forest in a horizontal belt at around 260 m (O. Tenow, personal observation, 1 July 1963). Hence, the population reached a maximum at an intermediate altitude in 1963, separated from the peak at the highest altitudinal station (350 m) in 1965. Later in 1965 defoliation confirmed the up-slope ‘spread’ (O. Tenow, personal observation, 11 August 1965).
In 1994–96, the peak density years of the O. brumata population at Gratangen also showed a significant interaction effect of year × altitude (d.f. = 4, 45, F = 10·33, P < 0·0001). The density peaked first in 1994 at an intermediate altitude (310 m) then declined in 1995. Densities at altitudes below increased during the same years (t-test of estimates 1994 and 1995, 120 m vs. 310 m and 230 vs. 310 m, d.f. = 45, t = −3·49, P = 0·0011 and t = −5·24, P < 0·0001). Between 1995 and 1996 the population densities at all altitudes increased, reaching the highest densities in 1996 in the lower part of the slope, i.e. altitudes < 390 m [general linear model (GLM) d.f. = 3, F = 67·08, P = 0·000]. This pattern is similar to that in the 1960s. At Engmo, the significant interaction year × altitude effect (d.f. = 6, 60, F = 4·34, P < 0·001) for 1994–96 reflects the modest density fluctuations at the low and intermediate altitudes (135, 200 and 260 m) in contrast to the density increase at the uppermost altitudinal station (350 m) reaching a high peak in 1996 (GLM altitudes 1996, d.f. = 3, F = 42·56, P < 0·001, Tukey's test 350 m vs. altitudes 135, 200 and 260 m, P < 0·04 or less) that collapsed in 1997. Also in this case the same pattern appeared as in the 1960s. In the 1990s, as in the 1960s, the succession of peak densities at different altitudes was the reverse of that at Gratangen. At both localities, these shifts implied vertical lags in population peaks (Fig. 5a,b).
In both periods, the E. autumnata population built to medium to high densities only in the upper part of the slopes. The populations peaked and collapsed 1 year earlier at Engmo than at Gratangen (Fig. 5c,d).
The 30-year population series of E. autumnata and O. brumata population dynamics at Budal exhibit peaks in 1975–77, 1986 and 1996–97 (Hogstad 2005). Only in 1997 did the O. brumata population reach outbreak density, with partial defoliation of the forest (Hogstad 1997). On average, O. brumata lagged 1 year after E. autumnata (Fig. 6, CCF-plot; systat 1992).
synchrony among regions
Data on caterpillar population densities were acquired with different methods at localities in the north and at Budal in the south. Furthermore, northern data were averaged over several altitudinal levels, whereas at Budal they were from one level. Nevertheless, a comparison of timing of population density fluctuations may be performed. The timing of density peaks (Fig. 2) indicates that the O. brumata and E. autumnata populations at Budal lagged 1 and 2 years, respectively, behind those at Frihetsli which, like Budal, is an inland locality. The same applies to a comparison with Engmo and Gratangen.
presence of synchrony
The combined outbreak area of E. autumnata and O. brumata in northern Fennoscandia has a wide east–west extension, i.e. from northern Finland, over northern Sweden and Norway where it extends into the Ofoten–Vesterålen–Lofoten island chain. Over a large part of this region, local outbreaks occurred in 1994–96. However, at the extreme north-western coast, fluctuation peaks occurred at the decadal shift 1999/2000 (marked 98 and 98–00 in Fig. 3). We argue that these peaks constituted the western endpoint of a wave of outbreaks moving from east to west over the period 1991–2000. Within the same time-period, side-waves seem to have moved northward. Contemporaneously, outbreaks moved from northern Fennoscandia, down the Scandes to southern Scandes (Nilssen et al. 2007; cf. Tenow 1972). Accordingly, within the east–west wave, local outbreaks in northern Fennoscandia were asynchronous. Similarly, outbreaks in the north were asynchronous with those in southern Scandes. Nevertheless, local outbreaks or peaks occurred successively in certain directions over large areas.
In their study of the importance of dispersal for the synchronization of O. brumata (and E. autumnata) outbreaks, Ims et al. (2004) quantified population density changes in 1999–2002 at six pairs of one large and one nearby (< 10 km apart) small outer island at the coast of north-western Norway north of the Ofoten–Vesterålen–Lofoten chain. The authors recognized peaks maximally out-of-phase (i.e. around 1995 and 2000) and concluded that this asynchrony occurred mainly between populations on small and large islands separated by open sea as a barrier to dispersal. In concluding this, Ims et al. (2004) questioned the large-scale spatial synchrony of E. autumnata and O. brumata outbreaks in general.
When interpreting these different peaks, the east–west moving wave of outbreaks must be considered. During their 4-year study, E. autumnata populations were low and were therefore omitted from the analyses. Of the studied O. brumata populations, four peaked in about 2000, three of them on the large islands in the south-west and the fourth also on the small island in the south-westernmost pair of islands. In the remaining eight populations to the north-east and east, fluctuations were low and indifferent. Our interpretation of this spatio-temporal fluctuation pattern is that the north-western coastal area became a zone of transition between the east–west moving wave which finally reached the outer west coast in 2000 and populations in the north-east and east that culminated around 1994–96. Indifferent fluctuations between about 1995 and 2000 may be ascribed to the transitional nature of the area. Thus, the findings of local population peaks maximally out-of-phase in northern Fennoscandia (or whole Fennoscandia) does not disprove the evidence of large-scale spatial synchrony of E. autumnata and O. brumata outbreaks (Tenow 1972; cf. Klemola et al. 2006). Instead, they reflect the existence of a wave (outbreak period XV). A similar ‘asynchrony’, caused by east–west moving outbreaks, is also indicated for the outbreak period XIII (Nilssen et al. 2007).
synchrony among localities
Fluctuation peaks as average for slopes occurred the same year (Fig. 2a) at relatively close localities, at Frihetsli, Engmo and Gratangen in 1994 (E. autumnata) and 1996 (O. brumata) and at Løbergsbukta and Melbu in 1998 (both species). Further, the fluctuation pattern was most associated at distances up to 60 km (Fig. 4). The tendency for an increasingly negative association at larger distances may reflect a temporal lag in fluctuations that at large distances results in fluctuations maximally out of phase. This can be interpreted as the result of a wave.
It is probable that the most frequently documented space/time pattern in ecological data is intraspecific synchrony in fluctuations of locally disjunct populations (Peltonen et al. 2002). Bjørnstad et al. (1999) identified three primary mechanisms for the rise of such a synchrony: dispersal among populations, dependence of population dynamics on a synchronous exogenous random factor such as temperature or rainfall (the ‘Moran effect’) and trophic interaction with other organisms that are themselves spatially synchronous (e.g. food) or mobile (e.g. parasitoids).
Parasitoids may travel great distances by wind, which may imply trophic interactions. Quality improvement of birch leaves as food in post-seed mast years has been suggested as a synchronizing force in E. autumnata outbreaks (Selås et al. 2001). This hypothesis was contradicted by feeding experiments (Klemola et al. 2003; but cf. Selås 2006).
The neonate caterpillars of E. autumnata and O. brumata may balloon (cf. Barbosa, Krischik & Lance 1989) several hundred metres (O. brumata, Edland 1971). According to genetic mapping, E. autumnata female moths can disperse over some kilometres (Snäll et al. 2004). The O. brumata female, with stunted wings, should be more spatially restricted. Only a small amount of dispersal between locally cyclic populations may cause synchrony (Barbour 1990; Bjørnstad et al. 1999). Moth dispersal in E. autumnata and larval ballooning in both species may therefore be important for a small-scale spatial synchronization of outbreaks.
There is a high overall synchrony in regional-scale weather variables which declines with distance in a similar manner to that in many animal populations (Koenig 2002). Thus, weather events may play an important role in synchronizing populations (Koenig 2002). In northern Scandinavia, winter and early summer temperatures show a high degree of synchrony over the region (Yoccoz et al. 2002). Winter temperatures might therefore synchronize outbreaks (Tenow & Bylund 1989). However, air temperatures at the coast never fall to levels lethal to either E. autumnata or O. brumata eggs (see ‘Historical documentation’) and therefore should not be of importance in coastal regions (Ims et al. 2004).
Often, synchrony in insect outbreaks declines more steeply with geographical distance than correlations in weather variables (Peltonen et al. 2002). It is probable, therefore, that local variation in habitat quality counteracts synchronization (Peltonen et al. 2002; Zhang & Alfaro 2003). The importance of local characteristics is illustrated by Engmo and Gratangen, where different patterns of outbreak progress of O. brumata and E. autumnata repeated almost exactly in two outbreak periods separated by 30 years.
Summarizing, neither ballooning by caterpillars nor low air winter temperatures may synchronize outbreaks other than over limited areas. Similarly, dispersing moths may synchronize fluctuations over some distance but probably cannot explain regional synchronization of outbreaks (see also ‘Synchrony among regions’).
synchrony among species
Expressed as means for Frihetsli, Engmo and Gratangen, the population development of O. brumata lagged 2 years after E. autumnata. At the two coast localities (Løbergsbukta, Melbu), E. autumnata and O. brumata peaked during the same year (low population levels). Instead, A. aurantiaria lagged behind, indicating that in increasingly diverse guilds time-lagged fluctuations should be anticipated (Fig. 2a). Also at Budal, about 700 km SSW, the O. brumata population lagged on average 1 year after E. autumnata (Fig. 6). Within the individual periods, there was a time difference of 1 year between peaks during the first and third outbreak periods, whereas no time-lag occurred during the second period (Hogstad 2005).
Interspecific synchrony has been demonstrated for several groups in a number of animal taxa (Liebhold et al. 2004), e.g. outbreaking oak-feeding Lepidoptera, among them A. aurantiaria and O. brumata (Raimondo et al. 2004) and tetraonid birds (e.g. Ranta et al. 1995). In contrast, aside from prey/predator and host/parasitoid relationships, we know of only one previous report on fluctuations synchronized with a time lag, i.e. on Canadian forest lepidopterans (Miller & Epstein 1986).
Hypotheses on interspecific synchronization include effects of common stochastic factors (‘Moran effect’) and trophic interactions where species have a common predator or share a fluctuating food resource (Liebhold et al. 2004). Hawkins & Holyoak (1998) have given an example of weather-caused simultaneous crashes, potentially synchronizing insect populations of widely different feeding groups over vast areas.
The 9–10-year E. autumnata cycle seems to be driven by delayed density-dependent factors, i.e. larval parasitoids and an induced resistance in birch leaves (e.g. Bylund 1995; Ruohomäki et al. 2000; Tanhuanpääet al. 2002) and possibly a cytoplasmatic polyhedrosis virus, lethal to caterpillars (Tenow 1972). O. brumata caterpillars share food with E. autumnata. The role of parasitoids for synchronization is not known, but several larval parasitoids are common to both species (e.g. Wylie 1960; Ruohomäki 1994; Kerslake et al. 1996). Several predators may be in common but their role as synchronizers is unknown. A polyhedrosis virus of the same type as in E. autumnata has been found in O. brumata caterpillars (Tenow 1972). Thus, some prerequisites for a synchronization of the two species exist. Nevertheless, a time-lag occurred between population peaks at four localities (Frihetsli, Engmo, Gratangen, Budal) and a synchronization at only two (Løbergsbukta, Melbu).
A pattern emerges: when population densities of E. autumnata and O. brumata were low, fluctuation peaks were synchronous (Løbergsbukta, Melbu: the second peak at Budal), at intermediate population densities (O. brumata at Frihetsli: the first period at Budal) and during outbreaks (Engmo, Gratangen: the third period at Budal) a time-lag developed between peaks. The increases of O. brumata occurred contemporaneously with E. autumnata population decreases. We hypothesize that the repeated decrease/increase courses point to some mutual influence(s).
Outbreaks of E. autumnata and O. brumata have been described as synchronous within the different outbreak periods (Tenow 1972). Now, the temporal relationships between the two types of outbreaks appear more complex. Many insect species oscillate with similar periods and a coincidence of peaks is likely to occur by chance (Johnson et al. 2005). However, if the time difference between the same phase in two cycling populations is constant (zero or more), the populations are ‘phase locked’ in synchrony (Lloyd & May 1999). Peaking occurred with time-lags varying from 0 to 2 years, possibly shifting with population densities. Thus, the fluctuations were not locked with a constant phase difference. None the less, the occurrence of the east–west wave in the north and the north–south wave along the Scandes, common to both species in the 1990s, indicates strongly that the synchronization of the two species was not coincidental. Considering also the repetition of common outbreak waves along the Scandes during the latest 11 outbreak periods (Tenow 1972; Nilssen et al. 2007), the information shows clearly that the two species’ cycles are coupled synchronously.
synchrony among regions
O. brumata and E. autumnata fluctuations at Budal, southern Scandes, lagged some years behind those at the northern, inland localities (Frihetsli, Engmo and Gratangen). This is in accordance with the finding that local outbreaks during outbreak period XV (1991–2001) moved with successive delays from northern to southern Scandes (Nilssen et al. 2007).
The outbreak waves of E. autumnata and O. brumata along the Scandes have moved from north to south, sometimes from south to north (Tenow 1972; Nilssen et al. 2007). Between these two ‘extremes’ there have been transitions when local outbreaks were synchronized to the same few years all over the Scandes, e.g. the concurrent outbreaks during the 1954–56 period that culminated all over the Scandes in 1955 (Tenow 1972).
The waves along the Scandes and the east–west wave in the north seem similar to the ‘travelling waves’ of outbreaks of the lepidopteran Z. diniana in the European Alps (Bjørnstad et al. 2002). Theoretical models of this system (Bjørnstad et al. 2002; Johnson et al. 2006) demonstrated that dispersal and gradients in favourable habitats could generate waves. Other quantitatively analysed travelling waves also seem influenced by the degree of landscape heterogeneity (tetraonid birds: Moss et al. 2000; voles: Lambin et al. 1998).
Both sexes of Z. diniana are strong flyers, capable of traversing vast mountain areas (Baltensweiler & Fischlin 1988). In comparison, E. autumnata and (particularly) O. brumata are poor flyers (see ‘Synchrony among localities’) living in a similar large-scale landscape with obstacles such as mountain ridges, barren areas above the forest limit, areas where egg-killing cold air accumulates in winter (Tenow 1975; Tenow & Nilssen 1990) and areas of young forests (see Introduction), lakes, fiords and straits (Ims et al. 2004).
Dispersal is always distance-dependent, which is why the level of synchrony decreases with distance (Ranta et al. 1995). However, the synchronization of the 1954–56 outbreak period did not decay over the approximately 1700 km NNE–SSW extent of the Scandes. Accounting for the formidable geographical obstacles to dispersal, the lack of large-scale spatial decay in synchrony during some outbreak periods and that interspecific synchrony cannot be explained by dispersal (Liebhold et al. 2004), we argue that dispersal has not contributed significantly to the large-scale synchronizations of the E. autumnata/O. brumata outbreaks. Sunspot activity has been suggested as an ultimate synchronizing factor (Ruohomäki et al. 2000; Selås et al. 2004). This hypothesis has been questioned seriously by Nilssen et al. (2007). Instead, interaction between regional stochasticity and variation in trophic interactions possibly explains both waves and contemporaneous outbreaks all over the Scandes (cf. Liebhold et al. 2004). However, small-scale waves due to dispersal may occur horizontally as well as vertically in valley sides (cf. Tenow, Bylund & Holmgren 2001; the present study).
We have used the expression ‘outbreaks spreading as a wave’ (Tenow 1972) because of the lack of a formal statistical analysis of travelling waves. Meanwhile, Klemola et al. (2006) have analysed spatio-temporal patterns in long-term series of E. autumnata fluctuations at localities all over Fennoscandia. Data indicate a potential travelling wave from NE/E to SW/W. This wave was an average for northern highly cyclic populations, and for several outbreak periods with outbreak waves moving along the Scandes at different speeds (Klemola et al. 2006: Fig. 3). For the part of the Scandes Fennoscandian, such an average appears to be the result of two components: the E–W- and the N–S-directed outbreak waves, as documented here and in Nilssen et al. (2007).
Britta, Christian and Lina Tenow, Henrik Nordenhem and Staffan Karlsson assisted in the field and in the laboratory. The Royal Swedish Academy of Sciences and Abisko Scientific Research Station offered lodging and working facilities. M. Sonesson, the former head and T.V. Callaghan, the present head of the Research Station, kindly supported the study. Two anonymous reviewers constructively criticized a first version of this paper. The Swedish Environmental Protection Agency (O. Tenow and A.C. Nilssen, project contract 27412) and the European Union (through the project HIBECO: H. Bylund, project contract QLRT-CT99-1515) funded the project. We thank all these people and institutions.