The effects of deer browsing on woodland structure and songbirds in lowland Britain

Abstract

The effects of deer in woodlands are known to result in habitat changes which can be detrimental to songbirds. In the first part of the paper we review the effects that deer may have on critical resources for woodland birds. The principal mechanism by which deer may affect habitat quality is through the reduction of low woody vegetation, which forms a key element of the preferred habitat of several species – this may be associated with loss of nest-sites, increased exposure to predators and reduction of food. The second part of the paper presents new evidence for the impacts of deer on vegetation structure, and how they may be contributing to the declines of some woodland songbirds in Britain. Results from extensive studies of deer at 13 woodland sites in England and Wales reveal that understorey foliage density decreases significantly with increasing deer density. Reduction of foliage was greatest at below 1.5 m above ground. We also report preliminary results of an experimental comparison of songbird densities between deer-fenced and unfenced areas of young coppice woodland in eastern England. In this study, deer browsing caused extensive impacts on vegetation structure in young coppice, though there was considerable variation between plots. Deer browsing resulted in reduction of canopy cover, reduction in density and cover of understorey vegetation, and an increase in grass cover. Abundance of bird species using the understorey, including all migrants, was significantly higher in coppice where deer were excluded. These results lend support to the hypothesis that deer are at least partly responsible for causing declines in some British bird populations, but they do not eliminate the possibility that increased shading is also responsible for changes in woodland structure. At present there is much spatial variation in deer densities in Britain so that impacts on low vegetation and habitat quality for birds are not expected to be the same in all regions. However, with continuing increases in deer numbers it is to be expected that such impacts will become more widespread.

Results from recent monitoring surveys have indicated that populations of several species of woodland birds are declining (Fuller et al. 2005, Hewson et al. 2007). The reasons for the declines are not entirely clear, and it is possible that a combination of factors is involved. However, there are increasing indications that deer may be responsible for causing a deterioration in habitat conditions for some woodland birds and Fuller et al. (2005) suggested that intensified deer browsing could potentially affect nearly a half of the declining species. Deer populations continue to increase in Britain (Gill 1990, Ward 2005) and they can cause substantial changes to woodland vegetation. Several of the woodland bird species about which there has been recent conservation concern depend on the understorey (defined here as all low woody and herbaceous vegetation within 2 m of the ground), the vegetation zone most directly affected by ungulate browsing. These species include Hedge Accentor Prunella modularis, Common Nightingale Luscinia megarhynchos, Song Thrush Turdus philomelos, Garden Warbler Sylvia borin, Willow Warbler Phylloscopus trochilus, Willow Tit Poecile montanus, Marsh Tit Poecile palustris and Common Bullfinch Pyrrhula pyrrhula. In contrast to these species, Wood Warbler Phylloscopus sibilatrix and Common Redstart Phoenicurus phoenicurus tend to be more abundant in open grazed woodland and may benefit from the impact of deer.

The effects of deer on woodlands in lowland Britain was the subject of a conference in 2000, which resulted in several reviews covering the effects of deer on woodland vegetation and fauna (Fuller & Gill 2001, Fuller 2001). It was clear from these accounts that deer browsing on vegetation has potential effects on the habitats, and consequently the populations, of a wide range of fauna, including invertebrates, small mammals and birds (Feber et al. 2001, Flowerdew & Ellwood 2001, Fuller 2001, Perrins & Overall 2001, Stewart 2001). However, there was a lack of direct evidence linking changes in British bird populations to deer as well as for the potential mechanisms (reductions in food sources or increased nest predation) involved. Since then, more evidence has become available, both in published sources and from our own work.

This paper is presented in two parts. First, we review evidence for the effects of deer on habitat structure and resources for birds. We also summarize evidence linking deer directly to bird abundance and breeding success. Secondly, we present new information, revealing evidence for the relationship between deer density and understorey vegetation density, as well as for the effects on songbirds of exclusion of deer from coppiced woodland. The emphasis of this paper is on woodland in the lowland parts of Britain where the typical deer species undergoing population increases are Roe Deer Capreolus capreolus, Fallow Deer Dama dama and Muntjac Deer Muntiacus reevesi.

REVIEW OF EVIDENCE

The effects of deer on woodland understorey habitats

The principal way in which deer are likely to affect birds is through changes in understorey vegetation structure caused by feeding (Fuller 2001). This could affect birds through changes in food supply, loss of nest-sites, increased vulnerability to nest predation (Martin & Joron 2003) or reduced roosting cover.

Deer are browsing ruminants and have a varied diet including the shoots and foliage of shrubs, young trees, climbers, forbs and grasses. In lowland woodlands in Britain evergreen species, such as Bramble Rubus fruticosus, Ivy Hedera helix and Honeysuckle Lonicera periclymenum can form a very important component of the diet, particularly in winter when less alternative green food is available (Hosey 1981, Forde 1989). The net effect of this browsing is to reduce the density and abundance of trees, shrubs and climbers in the understorey. This outcome is evident from many studies using exclosures, which reveal that browsing results in a reduced density and height of seedlings and saplings, or reduced cover of shrubs and climbers when assessed from either a vertical or a lateral perspective (Tilghman 1989, Putman et al. 1989, Stockton et al. 2005, Gill 2006).

When feeding on trees and shrubs, deer typically select the distal portion of shoots or leaves, because these contain higher nutrient concentrations and less lignin and cellulose than other plant parts. Besides reducing height and growth, this can also delay maturation and reduce flowering, which has further implications for birds as discussed later. The fact that the lower parts of plants are less likely to be browsed means that only the smallest seedlings are likely to be killed directly by browsing. Nonetheless, repeated browsing reduces the ability of trees and shrubs to compete so saplings and coppice stools may be killed by sustained damage. Studies investigating the impact of deer at differing levels of shade have found that vegetation is most severely reduced by a combination of browsing and shade, suggesting the two factors have an additive effect on the understorey (Tilghman 1989, Van Hees et al. 1996).

Investigations of deer browsing almost invariably reveal a change in plant species composition. Amongst tree seedlings and saplings, the reduction in density is usually associated with a loss of diversity (Gill & Beardall 2001). There is considerable consistency in the susceptibility of different tree species to deer browsing, the most vulnerable species including oak Quercus spp., willow Salix spp. and Ash Fraxinus excelsior. Reduction of Bramble is typical of heavily browsed woodland. The depletion of trees and shrubs is usually matched with an increase in the cover of grasses, sedges, mosses, bare ground or less palatable species such as Bracken Pteridium aquilinum (Putman et al. 1989, Kirby 2001). Many herbaceous species are also affected by deer browsing (Cooke & Farrell 2001, Tabor 2002) and it appears to be the taller species that are affected most (Kirby 2001).

Besides browsing, there are at least three other ways in which deer have a strong influence on woodland ecosystems. By feeding selectively on nutrient-rich foliage, deer alter nutrient pathways and may reduce nitrogen concentrations available to plants (Pastor et al. 2006). Secondly, deer feed on fruit and seeds, dispersing the seeds of some species through their gut and on their coats (Malo & Suarez 1995, Gill & Beardall 2001, Eycott et al. 2004). Deer also have an impact by causing physical damage with antlers and by trampling. These activities can kill seedlings and saplings and trampling can help to maintain areas of bare ground along runways or in wallows (Hester et al. 2006, Hobbs 2006). At present it is still unclear what the relative strength of these effects on vegetation can be in comparison with the direct effects due to browsing. Nonetheless, it needs to be appreciated that the impact of large herbivores such as deer is mediated through several processes, and these may affect woodland ecosystems at different rates and spatial scales.

Potential effects on food resources for birds

A number of studies have shown that invertebrates can be reduced by the effects of browsing. The response depends on the niche occupied by the species or group of invertebrate, but appears to be most severe for phytophagous invertebrates using the vegetation zones most affected by ungulate browsing (Den Herder et al. 2004, Suominen & Danell 2006). Baines et al. (1994), for example, found that Red Deer Cervus elaphus grazing upland pine woods reduced the abundance of invertebrates, mainly lepidopterous larvae. Allombert et al. (2005a) found that both herbivorous and anthophilous invertebrates were severely reduced by Black-tailed Deer Odocoileus hemionus browsing, although numbers of predatory and detritivorous insects were also reduced to a lesser extent. Other studies have reported reductions in some Diptera and Coleoptera (Putman et al. 1989, Gómez et al. 2004). The abundance of web-building spiders has been found to be linked to vegetation structure and consequently reduced by the effects of Sika Deer Cervus nippon browsing (Miyashita et al. 2004). Phytophagous insects are most likely to feed on the upper portions of shrubs and trees, for similar reasons to deer, namely that the upper parts are more likely to contain actively growing tissue and more nutrients (Baines et al. 1994, Pollard & Cooke 1994). They may therefore be disproportionately affected by deer browsing.

An additional consequence of browsing is that it is likely to reduce the frequency of flowering and fruiting. This may occur either as a result of the loss of flowering organs themselves, or due to a reduction in the size or frequency of flowering plants, or a delay in maturation (Allison 1990, Hester et al. 2006). Reductions in flowering could potentially affect the bird community in two ways, by reducing both the number of fruits and the number of seeds available later in the season, or by reducing the number of anthophilous insects dependent on nectar or pollen.

Invertebrates living and feeding closer to the ground show a variety of responses to large herbivores. Grazing and browsing tend to reduce the litter layer (Persson et al. 2005), which in turn may increase the area of bare ground and alter the surface temperature regime. Mobile ground-foraging species such as ants and lycosid spiders have been found to increase in response to grazing or browsing, whereas other groups such as harvestmen (Opiliones), which require humid conditions, may decrease (Suominen & Danell 2006).

Evidence of large herbivore impacts on woodland birds

Published evidence of the effects of deer on birds can be grouped into three kinds: designed experiments, natural experiments and circumstantial evidence.

The two main examples of designed experiments relate to White-tailed Deer Odocoileus virginianus in North America. In both cases deer caused major changes in low vegetation structure with knock-on effects on bird communities. DeCalesta (1994) simulated four densities of deer by using enclosures containing deer. Breeding songbirds were counted after 10 years of grazing. There were no effects on ground-nesters or canopy-nesters, but the species richness and abundance of species nesting in the ‘intermediate canopy’ (0.5–7.5 m above ground) were negatively related to deer density. McShea and Rappole (2000) used pairs of exclosures and unfenced control plots. Mist-netting was used to assess bird usage over 9 years. Exclusion of deer resulted in increases in overall bird numbers. Increases were recorded in both intermediate canopy-nesters and ground-nesters, although species requiring an open understorey decreased. Both studies indicate that changes in intensity of deer browsing can result in shifts in the composition of bird assemblages.

The islands of Haida Gwaii, British Columbia, have offered an exceptionally elegant natural experiment. Since their introduction in the late 19th century, Black-tailed Deer have gradually colonized some islands, with the result that huge differences in the plant and animal assemblages have now become apparent between islands with or without deer (Allombert et al. 2005a, 2005b). The abundance of songbirds was 55–70% lower on islands with the longest history of browsing compared with deer-free islands. Browsing shifted the community composition from one where species dependent on understorey vegetation dominated to one where species not dependent on understorey vegetation dominated. Furthermore, predation of nests by corvids was greater where concealment was reduced due to browsing (Martin & Joron 2003).

Finally, two British case studies provide circumstantial evidence. These involve changes in populations of species that depend on low vegetation that coincide with increasing impacts on vegetation structure from deer. The first case study is Wytham Woods, Oxfordshire, where there was an exceptionally large increase in deer pressure from the 1970s to the 1990s (Perrins & Overall 2001). Over this period Bramble declined and grasses increased, probably as a result of both deer browsing and canopy closure (Morecroft et al. 2001). The bird assemblages also changed greatly during the same period with large reductions in species depending on low vegetation, especially warblers (Perrins & Overall 2001). These changes in bird numbers were generally more marked in Wytham Woods than in other woods in the region.

The second case study is Bradfield Woods, Suffolk, where deer pressure also increased greatly during the 1980s and 1990s. Unlike Wytham Woods, which is predominantly broadleaved high forest, Bradfield Woods is managed coppice, cut on approximately a 25-year rotation. In 1989 a policy was adopted of protecting newly cut coppice by the erection of dead hedges (brushwood fences). These usually reduce deer browsing for the first 2 years, allowing re-growth of the coppice, but not preventing deer from subsequently modifying vegetation structure by removing low growth. Fuller (2001) showed that coppice structure in the mid 1990s was considerably different from that in the mid 1980s and that this was almost certainly a consequence of intensified deer browsing. Work in Bradfield Woods in 1987, before the large increase in deer numbers, showed that overall density of breeding songbirds was highest in coppice of 3–8 years growth, a phase when vegetation structure was at its most complex (Fuller & Henderson 1992). Species especially dependent on these vegetation structures included Common Nightingale, several warblers and Hedge Accentor. One might predict therefore that changes in vegetation structure as a consequence of browsing might affect these birds. By the mid 1990s there had been a large reduction in Nightingales but no major changes in other species depending on dense undergrowth (Fuller 2001). However, by the 2000s large decreases were also evident in densities of Willow Warbler and Garden Warbler. These species are also decreasing nationally but at a lower rate than in Bradfield Woods (Table 1). Blackcap Sylvia atricapilla and Common Chiffchaff Phylloscopus collybita have increased nationally, but at a faster rate than in Bradfield (Table 1). An experiment has now been initiated (see below) to examine whether the recent history of intense deer browsing in Bradfield Woods is a causal factor in these population changes (see below).

Table 1.  Estimated numbers of territories of breeding migrant passerines in Bradfield Woods between 1987 and 2006. Estimated densities (territories/ha) in 3–6-year-old coppice are shown in parentheses.*

	1987	1995	2001	2006	Within-wood % change 1987–2001	Within-wood % change 1987–2006	National % change†
* Note that similar areas of coppice are cut each year in Bradfield Woods so that there has been continuity of habitat for species associated with each stage of coppice growth. † National percentage changes are for the period mid 1980s to mid 2000s derived from Hewson et al. (2007; the plots censused by the BTO). No national data are available for Nightingale.
Common Nightingale	17 (0.6)	3 (0.1)	7 (0.4)	4 (0.4)	–59 (–33)	–76 (–33)	?
Blackcap	25 (0.9)	24 (0.8)	24 (0.7)	31 (1.5)	–4 (–22)	+24 (+67)	+57
Garden Warbler	36 (1.4)	38 (1.5)	21 (0.8)	18 (1.2)	–42 (–43)	–50 (–14)	–26
Common Chiffchaff	14 (0.4)	21 (0.7)	19 (0.7)	15 (0.6)	+36 (+75)	+7 (+50)	+155
Willow Warbler	64 (2.5)	54 (2.5)	17 (1.1)	15 (1.6)	–73 (–56)	–77 (–36)	–74

RECENT EMPIRICAL EVIDENCE

The effect of deer species and density on low foliage density

Hitherto, relatively little information has been available for lowland environments revealing how the severity of impacts changes with increasing population density, and how the impacts of each species of deer might differ. In view of the differences in quantity and quality of forage required by each species, some differences in damage to plants would be expected in response to changes in deer population density.

To address this need, one of us (R.M.A.G.) surveyed 13 mature woodland sites, assessing both deer population density and understorey vegetation density in late winter or early spring 2002 and 2003. The woodlands were located in lowland England and Wales (Hereford, Gloucestershire, Oxfordshire, eastern Powys, Surrey and Norfolk) and selected to include sites covering a wide range of deer densities. Each woodland site was subdivided into 1–5 blocks with 1–11 stands per block, yielding 27 blocks and 114 stands in total. Deer population density was estimated for each woodland using distance sampling methods (Buckland et al. 2001) based on observations of deer obtained at night using thermal imaging (Gill et al. 1997, Mayle et al. 1999). Foliage density was measured by recording the visibility of a 0.5 × 0.5-m frame from 10 m at four compass directions at ten plots in each stand. This was repeated at successive 0.5-m height intervals between ground level and 3.5 m. Visibility was recorded in three ‘intensities’ (0, entirely visible; 1, partially obscured; 2, totally obscured) and expressed as a percentage of the maximum possible score (8). Canopy cover was recorded by estimating overhead cover in 5% cover classes.

The density of deer in woodlands ranged from 0.0 to 59.9 animals/km² and included sites containing Roe, Muntjac and Fallow Deer, and a few Red and Sika Deer in various combinations. To simplify analysis, the smaller species (Roe and Muntjac) were analysed separately from the larger species, which tend to range further outside woodland. The results indicate a marked reduction in foliage density with increasing deer density (Fig. 1; Table 2), reducing the foliage density score by up to 92% of the zero deer density score. The reduction was greatest (and statistically more significant) at the height at which deer forage (below 1.5 m), and greater for larger deer species (at a given density) than smaller species. The effects of deer were significant after including canopy cover as a variable in all the models. Canopy cover had a weak negative effect on foliage density (significant only between 0.5 and 1.0 m) and a positive relationship above 2.5 m, because at these heights the lower part of the canopy affected the foliage assessments.

Figure 1. Graphical representation of the relationship obtained between deer and foliage density. Horizontal bars indicate the foliage densities for successive heights (from 0 to 350 cm) predicted from the parameters obtained from the REML models presented in Table 2. (a) ‘High’ density (60 km⁻²) of Fallow, Red or Sika Deer; (b) ‘High’ density (60 km⁻²) of Roe or Muntjac Deer; (c) no deer

Table 2.  Parameter values obtained for the mixed models (using REML) linking deer population density, deer species composition and canopy cover to foliage density. Methods used to assess foliage density, canopy, deer density and canopy cover are described in the text. ‘Deer species group’ refers to either larger species (Fallow, Red or Sika Deer) or smaller species (Roe or Muntjac). Forest block was incorporated into these models as a random effect. A separate model was obtained for each height range. The results indicate a closer relationship between foliage density and deer density below 1.5 m, at deer browsing height.

Height above ground (m)	Model parameters								R^2 (%)
	Constant		Deer density		Deer species group		Canopy cover
		P		P		P		P
Note the R^2 are from the weighted regression that approximates the REML fit.
0–0.5	5.66	<0.001	–0.0264	0.010	–1.584	<0.001	–0.0065	ns	29.7
0.5–1.0	5.16	<0.001	–0.0231	0.021	–2.026	<0.001	–0.01534	0.030	35.1
1.0–1.5	3.60	<0.001	–0.0234	0.007	–1.452	<0.001	–0.0009	ns	30.3
1.5–2.0	2.63	<0.001	–0.0163	ns	–0.873	0.019	0.0080	ns	15.2
2.0–2.5	1.62	0.002	–0.0097	ns	–0.371	ns	0.0199	<0.001	9.8
2.5–3.0	1.03	ns	–0.0042	ns	–0.056	ns	0.0253	<0.001	9.0
3.0–3.5	0.879	ns	–0.0004	ns	0.108	ns	0.0261	<0.001	9.3

Experimental evidence from coppiced woodland in eastern england

As far as we are aware, the experiment in Bradfield Woods is the first to be conducted in Britain on the effects of deer browsing on local densities and habitat use by birds. Preliminary results are reported here for 3-year coppice growth. A split-plot design was used to test whether total exclusion of deer alters the fine-scale distribution of birds. The experiment consists of eight plots of uniform coppice age of mean size 1.1 ha (range 0.8–1.5 ha). Half of each plot was fully protected from deer browsing immediately after felling and removal of the cut underwood by the erection of a 1.8-m steel deer fence. The remaining area was surrounded by a dead hedge. The experiment therefore compared total exclusion of deer with short-term exclusion. For convenience, the steel-fenced (i.e. total exclusion) subplots are subsequently referred to as ‘fenced’ and the paired dead hedge subplots as ‘unfenced’. The subplots were sufficiently large to accommodate the territories of several individual songbirds. Although small, the plots are considerably larger than the management compartments in most similar coppiced woods. Deer activity and browsing pressure within the wood have remained consistently high throughout the experiment.

The plots were established over a 5-year period, 1999–2003. The experiment will eventually allow comparisons of vegetation and bird abundance in coppice from zero to eight or more summers of growth. Here we present results from each of the plots in those years when the coppice had completed three full summers of growth (i.e. coppice in its fourth summer of growth is year class 3). These data were collected between 2002 and 2006. Three-year growth typically has reached the stage supporting relatively large numbers of warblers and other species dependent on dense low vegetation (Fuller & Henderson 1992).

In each year since 1999, the distribution of bird territories was established through territory mapping, with up to 15 visits in the period April to early June (Bibby et al. 2000). Several of these visits were evening or nocturnal visits exclusively for mapping Nightingales. In early June each year, vegetation structure was assessed at four fixed points spaced throughout each of the 16 subplots. The methods of estimating vegetation structure are summarized in Table 3. All bird mapping and estimation of vegetation was conducted by R.J.F. Territories were apportioned to subplots on the basis of the proportional distribution of registrations following Fuller and Henderson (1992). Effects of fences on vegetation measurements were assessed by general linear models run in MINITAB (Minitab Inc. 1999) in which presence or absence of a fence was treated as a fixed effect and plot as a random factor. Abundances of species groups in fenced and unfenced areas were compared using Wilcoxon matched pair tests.

Table 3.  Results of general linear models for effects of deer fence on components of vegetation in eight pairs of fenced and unfenced coppice in 3-year growth. All vegetation estimates were made at four points in each of the 16 plots. Note that mixed models were used in which fence was fitted as a fixed effect and plot as a random factor

	F	df	P
* Bramble, grass and canopy cover estimates to nearest 5% for a 5-m radius. † Mean of total vegetation cover at 0.5, 1.0 and 1.5 m above ground estimated to nearest 5% for a 5-m radius. ‡ Vegetation density using the method of Fuller and Henderson (1992). Means of visual distance estimates made on the four cardinal compass bearings.
Bramble cover*
Fence	0.90	1, 7	0.375
Plot	1.46	7, 7	0.314
Fence × Plot	7.02	7, 48	<0.001
Grass cover*
Fence	18.52	1, 7	<0.01
Plot	0.65	7, 7	0.709
Fence × Plot	9.60	7, 48	<0.001
Canopy cover*
Fence	5.63	1, 7	<0.05
Plot	0.62	7, 7	0.731
Fence × Plot	6.90	7, 48	<0.001
Low vegetation cover†
Fence	39.60	1, 7	<0.001
Plot	3.40	7, 7	0.065
Fence × Plot	2.05	7, 48	0.067
Field layer density‡
Fence	0.92	1, 7	0.370
Plot	2.42	7, 7	0.133
Fence × Plot	5.06	7, 48	<0.001
Shrub layer density‡
Fence	15.99	1, 7	<0.01
Plot	3.50	7, 7	0.060
Fence × Plot	2.23	7, 48	<0.05

Exclusion of deer was associated with significant effects on all six measures of vegetation structure. For five of the vegetation variables there was a significant interaction between fence and plot, indicating variation between plots in vegetation responses to exclusion of deer (Table 3). For the sixth variable, low vegetation density, there was a strong effect of fence. Bramble cover, canopy cover, low vegetation cover, field layer density and shrub layer density tended to be higher within the fenced subplots. Grass cover, however, was higher outside the fences. In the case of Bramble and field layer, the effect of fence was only apparent in the interaction with plot.

The fenced subplots held significantly higher densities of all migrants, long-distance migrants and of species dependent on a dense shrub layer than the unfenced subplots (Table 4). In the case of resident densities there was no significant difference (Table 4). In 2001 the densities of Garden Warblers in middle-aged coppice were much lower than in 1987 (Table 1). By 2006, despite persistently low overall populations, their densities in middle-aged coppice were similar to those in 1987. This indicates that the species had become increasingly concentrated into the middle-aged coppice, possibly as a result of the fencing. Blackcaps and Willow Warblers also increased their density in middle-aged coppice between 2001 and 2006.

Table 4.  Median density (territories per ha) and range of densities, in parentheses, of songbird groups in eight pairs of fenced and unfenced coppice plots in 3-year growth. Significance levels calculated using Wilcoxon matched pair tests.

	Fenced	Unfenced	P
* Sylvia communis, S. borin, S. atricapilla, Phylloscopus collybita, P. trochilus, Luscinia megarhynchos. † S. communis, S. borin, P. trochilus, L. megarhynchos. ‡ All resident passerines excluding Corvidae. § Species for which density of the shrub layer was the single best predictor of local abundance in an earlier study on Bradfield Woods (Fuller & Henderson 1992): Prunella modularis, S. borin, S. atricapilla, P. trochilus, L. megarhynchos, Turdus merula.
All migrants*	7.60 (4.2–14.0)	2.60 (1.9–6.4)	<0.01
Long-distance migrants†	5.35 (2.6–10.0)	1.50 (0.9–3.8)	<0.01
All residents‡	5.90 (4.6–8.1)	4.50 (1.1–9.4)	ns
Low shrub species§	8.50 (4.1–16.7)	2.15 (1.2–6.3)	<0.01

These results confirm that browsing has large effects on vegetation structure in coppiced woodland and that these effects are evident even where temporary protection is afforded to the coppice. These vegetation impacts appear to lead to local reductions in densities of those bird species most closely associated with complex low vegetation structures. The findings of the experiment are consistent with the hypothesis that declines in migrant bird populations in Bradfield Woods are at least partly influenced by intensified deer browsing mediated by changes in vegetation structure.

DISCUSSION

The recent evidence obtained on deer impacts gives increased support to the hypothesis that deer are capable of causing deterioration in habitat conditions for several woodland songbirds. There is widespread evidence that understorey vegetation is depleted by deer and comparisons between fenced and unfenced blocks of coppice woodland have revealed significant differences in densities of both migrants and understorey-dependent species. Further, recently reported studies based in British Columbia have also established links between the effects of deer browsing on vegetation, invertebrate populations, nest predation and songbird densities (Martin & Joron 2003, Allombert et al. 2005a, 2005b).

However, there is much that remains unclear about how deer are affecting British woodland bird populations. Apart from our results using fenced enclosures in Bradfield Woods, direct evidence of a link between deer numbers and songbirds is still lacking. It is still not entirely clear to what extent recent declines in woodland songbird populations have been caused by deer or whether they simply coincide temporally with a general increase in deer numbers. Nonetheless, the Bradfield Woods results do suggest that, at least at a local scale, reductions in some species may be associated with increased pressure from deer. Across Britain there is currently much regional variation in the abundance of different deer species and in total deer density. It should be noted that Bradfield Woods lies in a part of eastern England that holds one of the highest deer densities in the country. It cannot be assumed therefore that such impacts will be evident in coppiced woods in all parts of lowland Britain. However, deer are spreading and numbers continue to increase in many regions so it is to be expected that such impacts will become more widespread.

Increased shade from the woodland canopy may confound the impact of browsing. Both browsing and shading have similar effects, namely to reduce understorey vegetation. Many mature high forest stands in Britain have been simultaneously affected by both processes in recent decades. It is therefore difficult to distinguish the effects of each of these factors using survey methods. It would, however, be possible to separate the effects in an experimental investigation.

Evidence from investigations of browsing by deer in woodland suggest that increasing the level of shade appears to compound the effect of browsing by reducing understorey vegetation still further.

The more light-demanding understorey species appear to respond to increased light levels by recovering even where browsing pressure remains unchanged and there is evidence that bird species using the understorey can increase after the woodland canopy is thinned (DeGraaf et al. 1991, Bell & Whitmore 1997, Matsuoka et al. 2001). This has led to the suggestion from a North American study that effects of deer browsing may be offset by intentional thinning (DeGraaf et al. 1991). However, thinning at the intensity carried out for conventional woodland management in Britain is usually followed by rapid canopy closure, and it may therefore be necessary to apply a very heavy thinning to achieve a sustained increase in understorey cover. Furthermore, increases in the understorey cover may benefit deer or attract an increased browsing pressure. Efforts to improve understorey cover for birds may therefore be most successful if carried out in combination with deer management. There is clearly a need to understand more about how browsing and shading interact to affect understorey structures within different management systems.

Acknowledgments

We would like to thank Alan Ockenden, Dave Rodgers and Alistair Wybrow for assistance with field survey work and the support of the Forest Enterprise and private landowners for permission to use their estates. The Suffolk Wildlife Trust and English Nature gave permission for the experiment to be conducted in Bradfield Woods National Nature Reserve and we are especially grateful to Pete Fordham, the site manager, for his support. We also thank Amy Eycott, Chas Holt and two anonymous referees for reviewing the manuscript and Geoff Morgan and Stephen Freeman for statistical advice.