Spatial synchrony in field vole Microtus agrestis abundance in a coniferous forest in northern England: the role of vole-eating raptors

Authors


Dr S. J. Petty, Craigielea, Kames, Tighnabruaich, Argyll PA21 2AE, UK (fax 01700 811755; e-mail StevePetty@compuserve.com).

Abstract

1. The regional synchrony hypothesis (RSH) states that synchrony in microtine abundance over large geographical areas is caused by nomadic avian predators that specialize on small mammals for food. This has proved a difficult hypothesis to test because experiments at an appropriate scale are almost impossible.

2. We used the decline of the most abundant, nomadic vole-eating raptors in an extensive conifer forest in northern England (Kielder Forest) as a natural experiment to evaluate their influence on synchronizing voles at different spatial scales. Field vole populations fluctuated on a 3–4-year cycle of abundance, similar to the periodicity in central Fennoscandia.

3. Over a 23-year period, the combined numbers and density of kestrels and short-eared owls significantly declined. If these raptors were responsible for synchronizing vole abundance, the decline should have been associated with a decrease in synchrony. We could find no change in synchrony during the period of the greatest decline in kestrel and short-eared owl numbers (1980–97).

4. In Kielder, vole abundance has been shown to change in a wave-like manner, with synchrony in the direction of the wave being 5–10-fold smaller than that reported in Fennoscandia. Tawny owls are sedentary and the most abundant vole-eating raptor in our study area, and might have an equalizing influence on vole abundance over smaller areas if they foraged in a density-dependent manner and responded functionally to increasing vole density. If this was the case, spatial variability in vole density should have been less in occupied than unoccupied owl territories, especially in years of low vole density when owls could take a larger proportion of the standing crop of voles. Even though tawny owls caught a significant proportion of the vole population, we could find no difference in variation in vole density between owl territories that were unoccupied, occupied with no breeding attempt, or occupied with a successful breeding attempt.

5. We conclude that the small-scale synchrony in field vole abundance is unlikely to be caused by avian predators. Instead, it is more likely to be related to the pattern of clear-cutting that has developed in Kielder, which restricts vole dispersal. If this assumption is correct, we would predict more widespread synchrony in vole abundance in first-generation forests when extensive areas are planted over short periods of time, and this is supported by anecdotal evidence. These conclusions indicate that foresters may be able to manipulate the spatial dynamics of voles and vole predators by varying patch sizes within forests.

Introduction

Identifying mechanisms responsible for spatial synchrony in the dynamics of populations is the focus of much recent research (reviews in Bjørnstad, Ims & Lambin 1999; Koenig 1999; Caldow & Racey 2000). Specifically, research has focused on the contribution of climatic factors (Grenfell et al. 1998), dispersal (Kaitala & Ranta 1998; Baillie et al. 2000) and patterns of predation (Ims & Steen 1990) to patterns of synchrony, often extending over large areas (Ranta, Kaitala & Lindén 1995). The topic is of both fundamental and applied interest (Ormerod & Watkinson 2000). Indeed, the extent to which populations are synchronized impinges on the viability of fragmented populations and on the dynamics of pathogens.

In boreal regions, the hypothesis that predation by mustelids is responsible for the periodic high amplitude fluctuations in vole abundance has gained much support (reviews in Stenseth & Ims 1993; Korpimäki & Krebs 1996; Turchin & Hanski 1997). Spatial synchrony in vole abundance may be a consequence of predation by wide-ranging raptors (the regional synchrony hypothesis; RSH; Ydenberg 1987) that can respond both functionally and numerically to increasing vole abundance without a time-lag (Galushin 1974; Korpimäki 1985; Korpimäki 1986; Ydenberg 1987; Korpimäki & Norrdahl 1991; Korpimäki 1991a,b; Steen 1995; Norrdahl & Korpimäki 1996).

Support for the RSH comes from both spatially explicit models of predator–prey interactions and from experimental observations. Ims & Steen (1990) used a model to show that nomadic avian predators could cause population synchrony if they killed approximately 25% of the summer prey population and responded numerically, without a time lag, to spatial variations in vole abundance. Steen (1995) recorded the mortality rates of radio-collared root voles Microtus oeconomus, in a cyclic population still at peak density after the density in neighbouring areas had declined. He found that predators accounted for 82% of the mortality, of which approximately half was due to avian predators, and concluded that predation by nomadic avian predators was sufficient to cause population synchrony. Norrdahl & Korpimäki (1996) attempted to test the RSH experimentally by removing nest sites of some avian predators in five (four in some years) 3-km2 areas, 4–15 km apart, over 4 years in western Finland. Subsequently, spatial variations in prey density were higher among predator-reduction areas than among similar-sized controls, which suggested that avian predators might contribute to spatial synchrony. Further studies demonstrating that reductions in nomadic predator density can lead to higher variation in vole densities across a landscape would provide useful collaborative evidence for the RSH, particularly if the density of nomadic predators was reduced over large areas.

In this paper we focus on the role of vole-eating raptors in synchronizing field vole Microtus agrestis L. abundance in an extensive conifer forest in northern England. Here field voles inhabit the grassy vegetation in young conifer crops and numbers fluctuate on a 3–4-year cycle (Petty 1992, 1999; Petty & Fawkes 1997; Lambin, Petty & MacKinnon, 2000), similar to those in central Fennoscandia (Hanski, Hansson & Henttonen 1991). However, vole abundance has been shown to change in a wave-like manner, with synchrony in the direction of the wave being 5–10-fold smaller than that reported in Fennoscandia (Lambin et al. 1998). We show that the most abundant nomadic vole-eating raptors have been in long-term decline, hence providing a natural experiment to investigate the role of these raptors in regulating vole synchrony. If these species are able to synchronize vole abundance over extensive areas, the RSH predicts that the decline should be associated with a decrease in spatial synchrony in vole abundance. We also investigated the ability of a sedentary vole-eating raptor to synchronize vole abundance over smaller areas.

Materials and methods

Study area

Kielder Forest is one of the largest man-made conifer forests in Europe, with a total area of 620 km2 of which 500 km2 is planted with trees (Petty, Garson & McIntosh 1995). It forms part of a much larger area of man-made conifer forest planted over the last 70 years in the border area between England and Scotland. Kielder Forest comprises largely Sitka spruce Picea sitchensis (Bong.) Carr. and Norway spruce Picea abies (L.) Karst. The forest is managed on a clear-cutting system with 40–60-year rotations, which over the last 25 years has created a mosaic of different-aged stands of trees. Clear-cuts range in size from 5 ha to more than 100 ha, with the smallest in valley bottoms. Our study was confined to the main part of Kielder Forest District, which lies in Northumberland.

In our study area, the most important habitat for field voles is provided on clear-cuts, which develop grassy vegetation 2–3 years after trees are felled. These remain suitable vole habitat for about 12 years, until newly planted or naturally regenerated trees start to shade out ground vegetation. Virtually no grassy vegetation suitable for field voles is left in spruce stands after they reach 15–18 years of age. Thus, clear-cuts provide ephemeral islands of suitable vole habitat within a matrix of older crops that are unsuitable for voles. Earlier in the study period, areas of new planting on treeless moorland would also have provided habitat for field voles. Prior to planting, these moorlands would have held low densities of field voles (Charles 1981). With the removal of grazing sheep and the cessation of burning associated with tree planting, vole numbers would have increased until the growing trees shaded out the ground vegetation (Petty & Avery 1990).

Raptor data

Abundance of nomadic raptors

During the study period, the most abundant nomadic raptors that feed largely on field voles were the common kestrel Falco tinnunculus L. and short-eared owl Asio flammeus Pontoppidan. Short-eared owls are the most nomadic, and in troughs between vole cycle peaks they are largely absent. Kestrels are less nomadic (Village 1990). In poor vole years some stay, but use alternative prey (birds and invertebrates), breed later and have smaller broods. In good vole years, immigrant kestrels can substantially boost the breeding population.

The location of territorial pairs of kestrels and short-eared owls were recorded annually during 1975–97 as part of a long-term study of merlins Falco columbarius L. (see the methods in Newton, Meek & Little 1986; Little, Davison & Jardine 1995). Numbers of pairs were determined from a combination of sightings of displaying pairs and recently fledged young, and nests located. Thus assessed, the pairs recorded underestimated actual numbers by an unknown factor and therefore provide a population index. However, assessments were done in a consistent manner annually throughout the study, and should therefore have reflected changes in abundance among years.

Tawny owl numbers, clutch size and predation rates

The tawny owl is the most abundant raptor in Kielder. It is highly sedentary and overcomes years with few voles by staying on territory, ceasing to breed and by utilizing alternative prey, such as common frogs Rana temporaria L. and birds (Petty 1999). The number of territorial pairs of tawny owls was determined during 1981–97 in a smaller study area in the centre of Kielder Forest (Petty 1992). The study area measured 175·8 km2 and was defined by a north–south, east–west rectangle encompassing all nest sites used during 1979–98. Most pairs bred in nest boxes (Petty, Shaw & Anderson 1994). Prior to laying eggs, each pair visited most potential nest sites in their territory. Signs from these visits included traces of down or small body feathers adhering to the entrance hole. These signs occurred up to 1 month before the first egg was laid. Closer to laying, a deep scrape was formed in the debris at the bottom of the chosen site, often with down and small body feathers around the edge. In years when pairs did not produce eggs, they still went through this process, creating a well-formed nest scrape. Using these signs, a territory was classified as occupied when a fresh scrape, with down and/or small body feathers, was found in at least one nest site. Territories were regarded as unoccupied when no signs of owls were found at any of the nest sites during the last 2 weeks of March and the first 2 weeks of April.

During 1981–98, all potential nest sites were visited in the tawny owl study area during March and April to count the number of eggs laid (range 1–5). Additional clutch size data were collected in 1979 and 1980 from part of the tawny owl study area. Some female owls deserted during laying or incubation and relaid at either another nest site or, rarely, in the same nest (Petty 1992). In these cases, only one complete clutch per territory per year was included in the analysis (details given in Petty & Fawkes 1997). Territories that were occupied (see above), but where no eggs were laid, were assigned a clutch size of zero. Analyses were based on 993 clutches (including zeros) during 1979–98.

An estimate of the numbers of field voles removed by a territorial pair of tawny owls and their offspring was calculated for a year when vole numbers were low and increasing following the method given in Table 1.

Table 1.  Calculation of tawny owl predation on field voles (adapted from Lindén & Wikman 1983)
  • * Lowe (1980) gives 54 g per day for captive owls; 10% added to this for wild birds.

  • A 30-day nestling period followed by 70 days in natal territory before dispersal.

  • Adult food requirements used, as chick requirements will be less than adults in the early nestling period but greater than adults in late nestling period. Fledglings assumed to have similar requirements to adults.

  • §Mean fledged brood size from occupied territories in low and peak vole years (adapted from Petty 1992).

  • ¶Adapted from data in Petty (1992, 1999; unpublished data).

  • **From Petty (1999).

The number of field voles eaten each year by a pair of tawny owls and their offspring was calculated as follows:
FV = ((CA + CY) PV)/WV
FV = the number of field voles eaten by a pair of tawny owls during the year
CA = total food consumption of a pair of tawny owls (60 g food per day per bird* × 2 × 365 days)
CY = total food consumption by young (100 days × 60 g food per day × 0·27 (low vole year) or 2·62 (increasing vole year)§
PV = proportion biomass of field voles in tawny owl diet, 0·40 (low vole year) or 0·80 (increasing vole year)
WV = average weight of field vole (28 g)**

Field vole abundance

Vole abundance was measured in the tawny owl study area three times a year from summer 1984 to summer 1998, on 14–20 clear-cuts that had been planted between 2 and 12 years previously (Petty 1992, 1999). In each vole sign index (VSI) area, 25 quadrats (25 × 25 cm) were searched for the presence (= 1) or absence (= 0) of fresh (green) heaps of grass clippings. Thus, scores for each of the above VSI ranged from 0 to 25. These scores were converted to vole density (voles ha−1) from a separate regression for each season derived from live trapping (Lambin, Petty & MacKinnon 2000).

During 1996–98, another series of VSI assessments was undertaken that used 147 sampling points within the tawny owl study area (MacKinnon 1998). The methods used were the same as above and are hereafter referred to as the VSI2 areas. There were several of these within each tawny owl territory. This enabled mean coefficients of variation (CV) of vole density per territory to be compared between all occupied and unoccupied owl territories in three springs with contrasting vole densities (1996–98), thus controlling for variations in habitat and prey density between territory groups.

Field vole habitat

We constructed annual coverages of habitat in Kielder Forest within a geographical information system (GIS, Arc/Info version 7.1.2) (ESRI 1997). Polygon boundaries of uniformly planted areas were derived from Forestry Commission (FC) 1996 1 : 10000 scale digital stock maps. These did not include internal rides < 10 m wide. Attribute data for polygons were obtained from a FC digital database containing silvicultural and management information, including tree species, planting year and rotation (first rotation is the planting of previously treeless moorland, hereafter called new planting; second generation is the planting of clear-cuts, hereafter called restocking). Habitat suitable for voles within planted areas was defined as young tree crops between 1 and 11 years after planting. We retrospectively mapped the extent of this habitat for each year from 1975 to 1996 using query tools within the GIS. Other field vole habitats, such as grass pasture and moorland, occur within and around the study area, but the nature and spatial distribution of these have shown little change with time (Petty, Garson & McIntosh 1995). These permanent open areas have been excluded in the following analyses because we were primarily interested in how the spatiotemporal distribution of ephemeral vole habitats (new planting and restocking) has influenced vole and raptor dynamics.

Results

Availability of field vole habitat

The distribution and type of field vole habitat changed within Kielder Forest during the period of study (1975–97). At the start of the study, most of the tree planting was on previously treeless moorland (new planting) and was averaging about 400 ha year−1 (Fig. 1a). New planting peaked in 1977 and thereafter declined, with the last taking place in 1990. The decline in new planting was to a certain extent compensated by an increase in planting on clear-cuts (restocking) (Fig. 1a). Restocking peaked at 870 ha in 1988, and over the last 5 years averaged about 380 ha year−1.

Figure 1.

Potential field vole habitat in Kielder Forest, Northumberland, UK during 1975–97. (a) The area (ha) of new planting and restocking annually, and (b) the total area (ha) of new planting and restocking aged 1–11 years available annually.

The total area of potential field vole habitat available annually was calculated as the sum of the area of new planting and restocking aged between 1 and 11 years of age. This amounted to 6620 ha in 1975, peaked at around 7400 ha in 1979, and then slowly declined to around 5000 ha in 1986, at which level it remained until 1996 (Fig. 1b). Therefore, during the study period there has been a loss of around 20% in potential field vole habitat.

Trends in the abundance of nomadic raptors

Numbers (indices of abundance) for kestrels and short-eared owls declined over the 23-year study (Fig. 2a). There was no significant autocorrelative component in autoregression models for individual species or both species combined (Table 2). However, there was a significant decline in numbers over time for kestrel, but not for short-eared owl. In short-eared owls, the decline was largely manifest as progressively smaller peaks in a population that in most years was 0–2 pairs (Fig. 2a).

Figure 2.

(a) The number of pairs of kestrels (●) and short-eared owls (○), and (b) the density of both species per 100 ha of potential vole habitat. Both graphs are based on data for Kielder Forest District (excluding Kershope) during 1975–97.

Table 2.  Autoregression (AG) models showing the significance of declines in diurnal vole-eating raptors in Kielder Forest, Northumberland, UK during 1975–97. The first three models (a–c) use raptor numbers (index) and the last model (d) uses raptor density per 100 ha of suitable vole habitat. Exact maximum likelihood tests were used in all models
 CoefficientSEtProbability
(a) Kestrel (number)
AG−0·1010·227−0·4430·663
Year−1·0760·141−7·639< 0·0001
Intercept
(b) Short-eared owl (number)
105·879
12·147
8·716
< 0·0001
AG0·0530·2230·2370·815
Year−0·1760·261−0·6750·507
Intercept
(c) Kestrel + short-eared owl (number)
19·245
22·480
0·856
0·402
AG−0·0950·223−0·4280·673
Year−1·2570·306−4·107< 0·001
Constant
(d) Kestrel + short-eared owl (density)
125·595
26·394
4·759
< 0·001
AG−0·1890·226−0·8340·415
Year−0·0150·005−2·775< 0·05
Constant1·5590·4553·430< 0·01

If nomadic raptors influenced vole abundance, the combined numbers of kestrels and short-eared owls would be more important than totals of individual species. The combined numbers significantly declined during the study (Table 2). However, this analysis does not take into account the decline in the availability of field vole habitat (Fig. 1b). To control for this, the annual number of pairs of kestrels plus short-eared owls was expressed as pairs per 100 ha of vole habitat (Fig. 2b and Table 2). This analysis showed that the density of nomadic vole-eating raptors had significantly declined during the study.

Nomadic raptors and large-scale synchrony in vole abundance

During 1984–98, field vole abundance fluctuated on a pronounced 3–4-year cycle (Fig. 3a). If cycles had become less synchronized through time as kestrels and short-eared owls declined, this should have been reflected in the increasing variability in vole density on VSI sites. There was no evidence that this had occurred in relative terms (CV on time, rs = 0·003, n = 43, P = 0·99) (Fig. 3b). However, relative variability was significantly negatively related to vole density (rs = −0·632, n = 43, P = < 0·01). After effects of density were removed, there was still no indication that CV of vole density had increased during the study (partial correlation coefficient =−0·094, d.f. = 40, P = 0·55).

Figure 3.

(a) Temporal trend in field vole density (voles ha−1 with SE bars) and (b) temporal trend in variability (CV) of field vole density. Both graphs are based on three assessments per year from June 1984 to June 1998 (43 in total) in Kielder Forest, Northumberland, UK.

One problem with this approach was that our vole data went back only to 1984 (Fig. 3a), whereas the decline in numbers of kestrels and short-eared owls started from about 1980 (Fig. 2a). To extend the time series back to 1979, we used CV of tawny owl clutch sizes as a surrogate for relative variability in vole abundance, as there was a highly significant relationship between tawny owl clutch size and spring vole abundance (Fig. 4a). Even then, there was no indication of a significant temporal trend in the CV of clutch size (rs = 0·214, n = 19, P = 0·38) (Fig. 4b). There was a significant negative relationship over the years between the CV of clutch size and mean clutch size (rs = −0·939, n = 19, P = < 0·01), but after these effects were removed there was no evidence that the relative variation in clutch size had changed over time (partial correlation coefficient =−0·248, n = 16, P = 0·32). Thus, we were unable to detect any temporal change in the spatial synchrony of vole abundance throughout our large study area.

Figure 4.

Relationships between (a) mean tawny owl clutch size and mean spring (March) vole density (voles ha−1) (y = 0·251 + 0·016x, r2 = 0·74 during 1985–98), and (b) CV of clutch size during 1979–98.

Tawny owls and small-scale synchrony in vole abundance

Even if highly mobile raptors had no influence on the spatial synchrony of vole populations, resident raptors could theoretically equalize vole densities within their home range if their predation rates were sufficiently high, particularly in years when voles were scarce. Therefore, we used the tawny owl and the VSI2 data to investigate synchrony in vole abundance over smaller areas during three springs (1996–98) spanning one vole cycle.

We hypothesized that if predation by tawny owls was having a synchronizing effect on vole abundance within their foraging range, then variability in VSI2 scores within territories should decrease with increasing levels of foraging. Point coverages were made of owl breeding locations (nest boxes) attributed with breeding histories for each year and VSI2 sample locations attributed with scores for each spring. For every territory in the study area, the CV of VSI2 scores was calculated within circular areas of 200, 400, 600, 800 and 1000-m radii centred on the nest site. The mean nearest neighbour distance between territory centres was 0·74 ± 0·03 (SE) km (Petty 1992). For vacant territories, sampling areas were centred on the nest box most used in the past. We compared three territory classes in order of probable increase in vole predation: (i) vacant but occupied in a previous year; (ii) occupied but with no breeding attempt; and (iii) occupied and successfully fledging at least one chick. There were no differences in variability of vole abundance among these three classes within territories or years (Fig. 5a–c). Thus, there appeared to be no evidence of a synchronizing effect on vole abundance of predation by territorial tawny owls within their probable foraging ranges.

Figure 5.

Relative variation (CV) in spring vole density for three categories of tawny owl territory (unoccupied ●; occupied but with no breeding attempt ○; occupied and successfully fledging at least on chick ▪) in four radii (m) in 3 years (a–c). Vole abundance was low in 1996, increased in 1997 and peaked in 1998.

We then checked to see if this finding agreed with theoretical predation rates by tawny owls on field voles. We estimated that in a low vole year, tawny owls removed over the year around 650 voles per home range, and 1700 in a year when vole numbers were increasing (using data in the Table 1). The difference was the result of tawny owls producing larger broods and their diet comprising a higher proportion of voles in peak vole years than in low vole years. Median home range size of tawny owls in part of the study area was 113 ha, as estimated from radio-telemetry (Coles, Petty & Thomas, in press). The mean area of vole habitat (suitable clear-cuts and other treeless habitats) within a 600-m radius (113 ha) of the nest, estimated from the GIS, was around 27 ha. To this needs to be added 10% for unmapped open space, giving around 30 ha of vole habitat per home range. In a low vole year, 195 voles ha−1 were produced (39 voles ha−1 prior to breeding with 156 voles ha−1 added to the population during the year). In the following year when numbers increased, 395 voles ha−1 were produced (57 voles ha−1 prior to breeding with 338 voles ha−1 added to the population during the year). These figures were based on population density estimates by capture–recapture methods on eight live trapping grids during 1996 and 1997 (MacKinnon 1998; J.L. MacKinnon, unpublished data). Thus in low vole years, tawny owls appear to have the potential to remove around 11% of the standing crop of voles, and 14% in high vole years.

Discussion

The decline of kestrels and short-eared owls

The two most abundant species of nomadic vole-eating raptors at the start of our study have declined so dramatically that, during the latest vole cycle peak (1997–98), kestrels were rarely seen and no short-eared owls were recorded (Fig. 2a). We discuss possible causes for the decline of these diurnally active species elsewhere (S.J. Petty, unpublished data). We suggest that they are linked to declines of both species over a much larger geographical area and to increasing predation by a recently established population of northern goshawks Accipiter gentilis L. We do not consider that the decline is due to clear-cuts providing poorer hunting habitat than areas of new planting. Charles (1981) gives densities of 50–130 voles ha−1 in young forest (new planting) in southern Scotland. These figures can be compared with mean densities of over 150 voles ha−1 (and local maximum densities of > 350 voles ha−1) in peak vole years on restocked sites in Kielder (Fig. 3a; Lambin, Petty & MacKinnon 2000). Short-eared owls build nests on the ground, and they will as readily nest on clear-cuts as on areas of new planting (Shaw 1995; S.J. Petty unpublished data). Thus, we consider declines in numbers of kestrels and short-eared owls are unlikely to be related to the change in the type of vole habitat available (Fig. 1).

Large-scale synchrony in vole abundance

The RSH states that predation by nomadic vole-eating raptors synchronizes vole abundance over large areas. This hypothesis has been tested by one experiment, which claims to support the hypothesis (Norrdahl & Korpimäki 1996). We believe there are some questions over interpreting results of that study as support for the RSH because: (i) the scale at which the experiment was undertaken was very small compared with the area over which the RSH is postulated to operate; (ii) the distance between the control and predator removal area was given as 4–15 km, with no indication of the spatial distribution of the five experimental plot pairs; (iii) nest boxes were used for Tengmalm’s owl Aegolius funereus L. and kestrel in the control area, which could have enhanced the breeding density of both species; (iv) no attempt was made to reduce the breeding numbers of two ground nesting avian predators (short-eared owl and hen harrier Circus cyaneus L.) in the treatment area, but the densities of both species were lower in all 4 years in the control areas than treatment areas, suggesting there were inherent differences between treatment and control plots. All of which highlight the difficulty of conducting large-scale experiments (Ormerod & Watkinson 2000).

We have used the decline of kestrels and short-eared owls to investigate the role of nomadic raptors in synchronizing vole abundance over large areas. The main advantage of this approach is that the ‘natural experiment’ has been undertaken over a scale more appropriate for testing the RSH, compared with the 3-km2 plots used by Norrdahl & Korpimäki (1996). The disadvantages are the lack of replicates and controls, which can lead to problems with confounding effects. However, because of the large-scale nature of rodent synchrony, it may be quite unrealistic to test the RSH experimentally, as this would require landscape-scale treatments and controls.

The main candidate for synchronizing vole abundance should have been the kestrel, because earlier in the study substantial numbers were present even in low vole years, unlike short-eared owls (Fig. 2a). Hence, kestrels might have had a sufficiently high intake rate to have some spatial density-dependent impact on voles. Even though larger numbers of kestrels and short-eared owls were present in high vole years, it is unlikely that their intake rates would have been sufficient relative to vole abundance to have much impact. Thus, the decline of short-eared owls should have had relatively less impact on voles, as they were absent in many years (Fig. 2a).

We could find no evidence that variability in vole density had changed during the period over which kestrels and short-eared owls declined. This indicated that nomadic avian predators had little or no impact on synchronizing vole abundance throughout our study area, even when these two species of raptor were abundant. This finding does not preclude the RSH operating in Fennoscandia. In Kielder, it is possible that even when nomadic raptors were at their highest densities, they were still too scarce to influence vole synchrony.

We have shown elsewhere that field vole populations in our study area exhibit population cycles similar in periodicity to those in Fennoscandia (Petty 1992; Lambin, Petty & MacKinnon 2000). However, the spatial domain over which cycles occur are smaller than those recorded in Fennoscandia. Lambin et al. (1998) demonstrated that this was further characterized by vole abundance changing in the pattern of a travelling wave. The distance over which vole abundance is synchronized is 5–10-fold smaller in the direction (72°) of the wave than from other studies on microtine synchrony in Europe (Ims & Steen 1990; Steen, Ims & Sonerud 1996). MacKinnon (1998) has recently shown that this pattern persisted over a much larger area than that investigated by Lambin et al. (1998). This finding is consistent with the premise that synchrony is not caused by nomadic avian predators in our study area; if it was, vole abundance should have exhibited wide-spread synchrony.

Small-scale synchrony in vole abundance and tawny owl predation

Unlike nomadic avian predators, tawny owls stay on territory throughout a vole cycle, and therefore have the potential to equalize vole densities over a smaller scale by density-dependent foraging, particularly in the troughs between cycle peaks. We tested this hypothesis over one vole cycle (1996–98), which included a year (1996) with low vole densities. If tawny owls foraged in a density-dependent manner, and were able to deplete ‘hot spots’ of high vole density, then vole densities should have shown less variation in occupied than in unoccupied territories. We found no evidence that variation in vole density differed between territories in these two occupancy categories, even when we split occupied territories into those where no breeding occurred and those that were successful (fledged at least one chick) (Fig. 5a–c). This suggested that tawny owl predation had little effect in synchronizing small-scale patterns in vole abundance.

Ims & Steen (1990) estimated that nomadic and non-nomadic raptors could cause population synchrony if they removed at least 25% and < 17%, respectively, of the prey population. Our estimated predation rates on field voles by tawny owls (which are probably too high) are near to the estimate for non-nomadic raptors. However, it is highly unlikely that tawny owls were responsible for the pattern of synchrony observed (Lambin et al. 1998), as they would have to remove 40–50% of the vole population to compensate for the virtual absence of nomadic raptors from the tawny owl study area in 1996–98.

Concluding remarks and management implications

If, as indicated by our results, predation by vole-eating raptors has no effect on synchronizing vole abundance, what is the cause of the wave-like pattern of synchrony present on our study area (Lambin et al. 1998)? In our study, vole populations are synchronized over a relatively small spatial domain, but this area is still large enough to encompass numerous clear-cuts. MacKinnon (1998) has shown that density-dependent dispersal by voles can account for synchrony within patches (clear-cuts), but he thought it was unlikely for dispersal to be solely responsible for synchrony among patches. In our companion paper (Sherratt et al. 2000) it is proposed that the cumulative effects of limited short-distance dispersal may be all that is necessary to generate synchrony, and that the domain over which this occurs may be regulated by the connectivity of vole habitat. Thus in Kielder, where habitat for field voles (clear-cuts) currently occurs as islands amongst otherwise unsuitable habitat (closed-canopy spruce forest), dispersal by voles is restricted and synchrony domains are smaller than in extensive areas of homogeneous vole habitat, such as in Fennoscandia (Sherratt et al. 2000), where there are fewer barriers to dispersal. This may not always have been the case in Kielder, and may be associated with how second-generation forests are managed. Anecdotal evidence indicates that vole abundance can be synchronized over much larger areas in first-generation forests, when thousands of hectares are often planted over a short period of time (Elton 1942; Lockie 1955; Charles 1956; Village 1990), or in extensive areas of grassland (Taylor 1994). This suggests that foresters may be able to manipulate the spatial dynamics of at least some species by varying patch sizes of tree stands within forests.

Acknowledgements

We thank Forest Enterprise for their help and for allowing us to work for so many years in Kielder Forest District; David Anderson for assisting with some of the fieldwork; and Professor Ian Newton and two anonymous referees for helpful comments on the manuscript. Some of this work was supported by a grant from the Natural Environment Research Council and the Scottish Executive Rural Affairs Department under the Large-Scale Processes in Ecology and Hydrology Thematic Programme (number GST/02/1218).

Received 22 January 1999; revision received 16 September 1999

Ancillary