A model of a natural wooded landscape in Britain as ...

A model of a natural wooded Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
landscape in Britain as influenced
by large herbivore activity

K.J. KIRBY

English Nature, Northminster House, Peterborough PE1 1UA, England
E-mail: [email protected]

Summary

The generally accepted view of the natural forests that once covered much of Britain has been of
largely closed-canopy woodland with regeneration in gaps created by the death or destruction of
small groups of trees, or occasional catastrophic blow-downs. An alternative view has recently been
promoted in which large herbivores grazed open areas that eventually went through scrub and
woodland phases. Regeneration within the woodland phase was prevented by grazing so that, when
the trees died, open areas formed again. For regions where the herbivore-driven process
predominated a model of possible landscape structures was developed for an area of 5 ϫ 5 km
(2500 ha) over a 500-year cycle. Different combinations of open and closed landscapes resulted,
depending on the assumptions made about how long, for example, the closed woodland or open
park phases lasted and whether patches of a particular phase were clumped or scattered. Possible
implications for different species groups are illustrated. The balance between open and closed areas
depends on assumptions made about the duration of the different phases. However, a predominantly
wooded landscape (50 per cent in the woodland phase, 25 per cent in the park phase, with the
balance as scrub (15 per cent) or break-up (10 per cent) stands) is compatible with a herbivore-
driven dynamic process. Continuity of both open conditions and old trees over time and space can
be achieved in this wooded landscape with an intimate mix of habitats at the scale of a few hundred
metres. Modern conservation priorities in Britain are set against a background of 3000 years (at
least) of a cultural landscape and controlled agricultural-type grazing is part of the management used
to maintain some priority habitats and species. Under modern conditions free-range grazing regimes
may not produce the same effect and so cannot, a priori, be assumed to be better in biodiversity
terms than controlled grazing in all circumstances. However, the model results suggest that over
large spatial scales and a long time period varied habitat mosaics may be generated by naturalistic
grazing regimes. Naturalistic grazing regimes should be trialled therefore to determine what range of
habitats and species they can now support and as a way of improving our understanding of how the
former natural landscape might have functioned. There are practical issues including animal welfare
and public liability that need to be resolved if such trials are to be taken forward.

© Crown copyright 2004. Published by Oxford University Press Forestry, Vol. 77, No. 5, 2004

406 F O R E S T RY

Introduction Vera (2000, 2002a, b) puts forward a case that, Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
across western Europe, large herbivores not only
A widely accepted view of the natural forests that created open conditions, but drove the whole
once covered much of western Europe is that they landscape regeneration cycle. Herds of aurochs
were largely closed-canopy woodland, with many Bos primogenius, bison Bonasus bonasus and
mature, shade-tolerant trees and shrubs; regener- wild horse Equus spp. grazed large open areas
ation was in gaps created by the death of small that eventually went through scrub and
groups of trees or occasional catastrophic blow- woodland phases. Grazing prevented regener-
downs (e.g. Peterken, 1996; Rackham, 2003). ation within the woodland phase, so that when
Only once humans started to become abundant the trees died open areas formed again (Figure 1).
would the landscape have become very open This is attracting attention in relation to the
(Buckland, 2002). management of large nature reserves such as the
Oostvaardersplassen in the Netherlands (Kampf,
The natural forest may, however, have been 2002a, b), but also with respect to large wood-
more open than is sometimes inferred from pasture sites such as the New Forest (Hampshire,
pollen evidence alone (Buckland, 2002), a view UK). Vera (2000) uses this large mosaic
shared by, for example, Tubbs (1996). Openness (~38 000 ha) of grassland, heath, mire and
might have been created by a variety of factors woodland, grazed by cattle and ponies (Tubbs,
including flooding, disease, extreme storms or fire 1986), as an analogue for some of the processes
(Whitbread, 1991; Green, 1992, 2002; White- involved in his cyclical turnover of vegetation.
house, 2000; R.H.W. Bradshaw, personal com-
munication). Rackham (2003) and Bradshaw et al. (2003)

2. Scrub phase: spread of 1. The open or Park phase: largely
thorny shrubs excludes open landscape with a thin scatter
herbivores; young trees grow of trees left from the previous
up with the shrubs and grove; vegetation mainly grassland
eventually overtop them. or heath species.

2. Scrub 1. Park

3. Grove 4. Break-up

3. Grove: tree 4. Break-up: period
dominated phase of the during which the canopy
cycle; closed canopy opens out as trees die;
shades out the shrubs; vegetation shifts from
herbivores return, woodland to grassland
preventing regeneration. species

Figure 1. Vera’s model, consisting of the three phases of open park, scrub and grove, to which a fourth
break-up has been added to represent the transition from woodland grove back to open habitats (park).

LARGE HERBIVORE INFLUENCE ON NATURAL WOODED LANDSCAPE 407

discuss the arguments used by Vera (2000) From the patterns generated by the model sug- Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
against the closed-canopy woodland landscape. gestions are made about the likely suitability of
However, if we accept that at least parts of the such a landscape for different groups of organ-
landscape showed a cyclical turnover of vegeta- isms. Some implications for modern conservation
tion, driven by large herbivores, what would that in a British context are discussed.
landscape look like?
Methods
Vera (2000) suggests, in making a link between
groves, ‘hagas’ and deer parks, that groves might Assumptions about changes over time
often be 40–80 ha or even, through analogies
with the New Forest, up to several hundred At a certain point in the landscape there would
hectares. Elsewhere regeneration is described as be grassland first, followed by thorny scrub or
being in small groups or as individual trees. The other unattractive (e.g. unpalatable) species of
extent of the different phases and how long it plant, then forest (grove) and finally back to
would take for any patch of land to go through grassland (Figure 1) (Vera, 2000). However, at
the complete cycle have implications, not just for any one time, some points in the landscape would
the interpretation of evidence for and against his be at the grassland phase (hereafter referred to as
theory as applied to past landscapes, but also for ‘park’ to reflect the scattered survival of trees
its relevance to modern conservation. A hypo- through this open phase; see later), some at the
thetical landscape was therefore set up composed scrub phase, some as groves (Figure 2). These
of 1-ha cells, each of which moved through the phases would not be of equal duration.
Vera cycle (Figure 1). The time spent in the
different phases was varied, according to assump- The initial assumption was that it would take
tions about the time it would take for scrub to 500 years for any patch of ground to go through
develop and be overtopped by trees, and the life the cycle (Figure 2). This cycle was broken-down
expectancy of trees. Different ways in which the into twenty 25-year stages. Variations in the
open and closed phases might be distributed
across the landscape were explored.

500 year cycle divided into twenty stages of twenty-five years each, with the different phases in the cycle at each of five points

across a landscape.

Stage 1 26 51 76 101 126 151 176 201 226 251 276 301 326 351 376 401 426 451 476

start

year

Point Park Scrub Grove Break-up Park

a xxxxxx xxx

Point Break- Park Scrub Grove

b up xxxxxxxxxx xxxx

Point Grove Break-up Park Scrub Grove

c xxxxxxxxxx xxxx

Point Grove Break-up Park Scrub

d xxxxxxxxx xxxx

Point Park Scrub Grove Break-up

e xxxxxxxxx xxxx

Figure 2. Representation of the changes in the Vera cycle across time at different points in a landscape.
xxxxx = the regeneration period for each point.

408 F O R E S T RY

structure and composition of a patch within each The length of about 250 years for grove phase Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
25-year period were ignored for the purposes of is based on typical current life-spans for the
the model. The basis for choosing the initial majority of the trees of about 300–400 years and
phase lengths is discussed below. Regeneration of is consistent with the age of bog oaks.
scrub and trees is assumed to take place only
during the latter part of the relatively open, Break-up (50 years)
grazed conditions (Vera, 2000). Thus, in the
model, the regeneration phase has been set as Closed canopy stands of broadleaves may be
the last 50 years of the park phase and the first reasonably stable in Britain up to about 250
25 years of the scrub phase, 75 years in total, or years. Thereafter, particularly if dominated by
15 per cent of the whole cycle. beech (Fagus sylvatica) rather than oak (Quercus
spp.) they may break-up either through indi-
Scrub (75 years) vidual tree deaths or through catastrophic events
such as the 1987 storm (Whitbread, 1991;
Once scrub started to form it would spread fairly Spencer, 2002). Disease (Green, 1992; Dobson
rapidly as it increasingly hindered movement of and Crawley, 1994) could also contribute to
animals. Blackthorn (Prunus spinosa) scrub can stand break-up. The initial break-up event could
spread concentrically through root suckering, so happen overnight if it was caused by a severe
as to form a thicket of 0.1–0.5 ha in 10 years storm and the shift in the characteristics of the
(Vera, 2000). However, even where the main stand from grazed woodland to open grass-
scrub species was hawthorn (Crataegus spp.) land/heath could be rapid. This shift might be
establishment of new hawthorn bushes would be delayed if there were substantial amounts of
favoured close to existing ones leading to patch fallen branches and trunks that prevented access
formation because of higher seed availability and by the grazing animals (Green, 2002).
increased protection from grazing closer to
existing bushes. Once formed, scrub would be Not all the trees might be killed during break-
relatively short-lived as the dominant component up. A few are assumed to survive through the
in that patch, because trees would establish whole of the next park and scrub phases (another
within the protection of the scrub and grow up 200 years), before being shaded out by the next
with the bushes, eventually overtopping them. generation of trees in a grove.
Establishment of only two or three trees per
hectare per year during the regeneration window Park (125 years)
would be sufficient to form what was effectively
closed canopy woodland (Kirby, 2003). This open element of the landscape is one key
difference in Vera’s ideas compared with the trad-
Grove (250 years) itional gap-dynamic model – hence a relatively
long time has been allowed for it (equivalent to
The young trees growing up with the scrub several generations of large herbivores); other
would be subject to lateral shading from first the differences with the gap-dynamic model are dis-
shrubs and then their neighbours in the grove. cussed in Vera (2000), Bradshaw et al. (2003)
They would tend to be form tall maiden stems and Rackham (2003). Following Vera (2000) it
(unless they were at the edge of a grove), similar has been assumed to be predominantly open, but
to the New Forest ‘Generation B’ trees (Peterken allowance has been made for a thin scatter of
and Tubbs, 1965). Such tall, unbranched trees trees (20 per cent canopy cover at the start of the
from the Atlantic period have been dug out of period but declining to 2 per cent cover by the
the fens and bogs in Britain. One such was at end) because it seems unlikely that the transition
least 340 years old, and more than 1 m in between the different phases would be absolutely
diameter 6.5 m from the roots; others had clearly sharp. During this open period there could be
grown up, closely spaced with lengths of 27 m to some attrition of the main canopy of these sur-
the first major branch (Peterken, 1981; viving trees, because of increased exposure, but
Rackham, 2003). also re-iteration of the canopy from lower
branches. Such trees could be the equivalent of

LARGE HERBIVORE INFLUENCE ON NATURAL WOODED LANDSCAPE 409

‘veteran trees’ in modern wood-pastures (Read, way, but the variations reflect real ecological pro- Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
2000). cesses. For example, if animals concentrated their
grazing in particular areas then open park phase
Developing the pattern of spatial variation cells might be clumped. Soil fertility differences
or proximity to water might lead to some cells
The cycle of vegetation change was applied to a moving through the cycle more rapidly than
hypothetical area of 5 ϫ 5 km (2500 ha). Large others.
grazing animals may range several kilometres in
a day (Putman, 1986), and are assumed to have Scenario 1
access to all parts of the 5 ϫ 5 km area except All the 25-year age stages were represented by
insofar as vegetation structure limits their 125 1-ha cells. The cells allocated to a particular
movement. stage were distributed at random across the land-
scape. All cells moved through the cycle at the
The 5 ϫ 5 km area was broken down into same rate. In effect this reflects a homogenous
2500 1-ha cells. One hectare is large enough for landscape with potentially free access by the
a large grassland glade to be able to establish in herbivores to all parts of the landscape.
it during the break-up phase. In terms of light
availability such glades are also large enough to Scenario 2
permit oak regeneration. Mellanby (1968) noted This differs from scenario 1 in that the cells were
that, while young oaks did not grow directly clumped in blocks of 25 to give a coarser grain
under the canopy of an existing tree, there were to the landscape. Within any one 25-ha block the
over 60 within a few metres of the edge of its allocation of cells was restricted to one dominant
canopy. If a smaller cell-size were adopted there phase and the phases that preceded and followed
would be significant side-shading effect from the dominant phase. This represents a situation
adjacent trees, and open grassland and scrub where, perhaps because the animals tend to form
species would be less able to thrive in the gaps larger herds, at any one time their impact is more
created. concentrated in one part of the landscape; there-
fore if one cell is at the park phase, its neigh-
Developing the model: different landscape bouring cells are also likely to be open.
scenarios
Scenario 3
Each cell was assigned a unique set of coordinates The landscape was divided into three portions
in the 5 ϫ 5 km landscape in a row on a spread- (800 ha, 900 ha, 800 ha) from west to east. In the
sheet. Cells were assigned a starting point at one first portion the cycle length was shortened, the
or another of the 20 25-year stages in Figure 2. relative importance of the grove phase was
Each cell then moved through the Vera cycle in reduced while that of the park phase increased
these 25-year stages from its assigned starting (Table 1). The central part of the landscape had
point (e.g. points a–e in Figure 2). At any one the same phase lengths as for scenario 1. The
time in the cycle the extent of the different phases easternmost section operated with a reduced park
(park, scrub, grove, break-up) was calculated as phase and lengthened grove phase. This situation
to the number of cells that were in the relevant could represent the impact of heavier grazing
25-year stages that contribute to each phase. (hence a longer park phase), in proximity to (say)
a river, whereas the more distant part of the land-
Initially assignment to a starting stage for each scape had less grazing, and regeneration within
cell was at random and each 25-year stage had the park phase occurred sooner.
125 cells, a 20th of the landscape, assigned to it.
The rate of progress through the cycle was the Scenario 4
same for all cells. Different scenarios were then The ecological analogue for this scenario is that
developed in which the cells assigned to a particu- variability could exist in the landscape or grazing
lar 25-year stage might be clumped rather than pressure at the scale of the individual cell. Patches
scattered randomly; and the rate at which the that were favoured by animals might be held
cells moved through the cycle was also varied.

Each scenario was developed in a mechanistic

410 F O R E S T RY
Table 1: Phase lengths adopted in different sections across the landscape under scenario 3

Sections (from west to east) Phases Length of phase (years) % landscape in that section

‘Long park’ (800 ha) Park 200 40
Scrub 150 30
Grove 100 20
Break-up 10
50

‘Standard’ (900 ha) Park 125 25
Scrub 75 15
Grove 50
Break-up 250 10
50

‘Short park’ (800 ha) Park 50 10
Scrub 75 15
Grove 300 60 Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
Break-up 75 15

longer in the park phase; a small patch of infer- Under even progression, such as in scenario 1,
tile soil might scrub up more quickly than all cells would move on 20 stages over a 500-year
another; different cells might survive for longer period. The actual number of stages that cells
periods in the grove phase if sheltered from high moved on in this scenario is shown in Figure 3:
winds. values less than 20 represent cells taking longer
than 500 years to move through the cycle; values
In the model this variability was mimicked by more than 20 represent cells moving through the
moving individual cells through the cycle at cycle in less than 500 years. There are more of the
different rates. At each 25-year stage transition, a latter (i.e. faster cycling) because there were 1000
factor was introduced to determine whether the cells at each transition that could make either
cell simply moved to the next stage in an orderly two- or three-stage jumps forward, compared
progression (as for scenarios 1–3), stayed at the with only 500 where no transition occurred.
stage it was at (effectively lengthening the cycle for
that cell), or jumped two or three stages forward Scenario 5
(effectively shortening the cycle). At each stage, The assumptions were altered to produce a more
1000 cells were allocated the normal transition, open, savannah-type landscape. The length of the
while 500 each got either the no-change, two- grove phase was reduced to 100 years and there
stage or three-stage transitions. The transition was an earlier, but much longer break-up period
factors were re-randomized at each change, so the (175 years), during which the tree canopy gradu-
rate at which a cell moved through the cycle also ally declined from 80 per cent to 20 per cent
varied during the cycle. This process was repeated cover. Such early break-up might be triggered by
for the 20 successive stage-transitions. bark-stripping of the trees. The park phase was
left at 125 years; the scrub phase lengthened
Each cell must still go through all phases in slightly to 100 years.
sequence, except in one exceptional circum-
stance: if a cell (say row a in Figure 2) was at the The extent of different cover types and potential
end of the grove phase and jumped forward three habitat for different species groups
stages then the break-up phase could effectively
be missed. This is not, however, an impossible An initial criticism of the scenario modelling
situation: a patch could blow down in a couple was that it gave the impression that there are
of years, and if this were combined with a high sharply defined boundaries between the phases,
local grazing pressure at that point, the transition whereas in reality they would grade into one
from grove to park would (from the point of view
of the model) be more or less instantaneous.

LARGE HERBIVORE INFLUENCE ON NATURAL WOODED LANDSCAPE 411

Figure 3. Frequency of cells showing different cycle lengths. Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016

another. For example, scrub would be present in 2 The ground flora consisted of just two types of
the grassland before it was sufficient to class the species: those normally now associated with
area as part of the scrub phase; a few scattered grassland that tend to be reasonably tolerant of
trees could be present in the open park phase. high levels of grazing, but intolerant of dense
Therefore landscape structure and composition shade; and those now associated with
were also explored in terms of how the vegeta- woodland that tend be less tolerant of grazing
tion might vary across the landscape under a (Rackham, 2003), but more tolerant of shade.
given scenario, irrespective of phase type. For Grazing was assumed to be highest in the park
simplicity, just the ‘wooded landscape’ (scenario phase, to decline during the scrub phase and
1) and the ‘savannah’ options (scenario 5) were early grove phase and thereafter increase
considered. again. Light at ground level was assumed to be
high during the park phase, very low under the
The following assumptions were used to scrub phase and to increase during the latter
explore the potential distribution of different part of the grove phase (opening of tree canopy
groups of species. and loss of shrub layer).

1 Tree and shrub cover at a point followed the 3 Total ground flora cover, and whether of a
patterns in Table 2 (columns 4 and 5). woodland or grassland type, varied at different

412 F O R E S T RY

stages through the cycle as shown Table 2 distribution of such birds, e.g. some warblers
(columns 6 and 7). often now associated with coppice, was taken to
4 Many woodland birds are associated with scrub follow that of areas with more than 20 per cent
and woodland understoreys (Fuller, 1995). The shrub cover (column 9 in Table 2).

Table 2: Assumed changes in the cover of different components of the landscape for (a) scenario 1 and (b)
scenario 5

Year 25-year Phase Tree Shrub ‘Grass’ ‘Wood’ Veteran ‘Shrub
stage cover cover flora flora trees nesters’
cover cover present
present

(a) Wooded landscape, scenario 1

1 1 Park 5 0 98 2 Yes

26 2 Park 2 25 75 2 Yes Yes Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016

51 3 Scrub 2 50 50 2 Yes Yes

76 4 Scrub 25 100 5 5 Yes Yes

101 5 Scrub 50 100 2 50 Yes Yes

126 6 Grove 90 75 2 75 Yes Yes

151 7 Grove 90 50 2 98 Yes Yes

176 8 Grove 90 25 2 98 Yes

201 9 Grove 90 25 2 98 Yes

226 10 Grove 90 25 2 98 Yes

251 11 Grove 90 5 2 98

276 12 Grove 80 5 10 90

301 13 Grove 80 5 10 90

326 14 Grove 80 5 20 80

351 15 Grove 80 5 20 80

376 16 Break-up 70 0 25 75

401 17 Break-up 50 0 25 75

426 18 Park 20 0 50 50

451 19 Park 15 0 75 25 Yes

476 20 Park 5 0 98 2 Yes

(b) Savannah, scenario 5

1 1 Park 5 0 98 2 Yes

26 2 Park 2 25 75 2 Yes Yes

51 3 Scrub 2 50 50 5 Yes Yes

76 4 Scrub 25 100 5 5 Yes Yes

101 5 Scrub 50 100 2 10 Yes Yes

126 6 Scrub 75 75 2 10 Yes Yes

151 7 Grove 90 50 2 50 Yes Yes

176 8 Grove 90 25 2 98 Yes

201 9 Grove 90 25 2 98 Yes

226 10 Grove 80 25 2 98 Yes

251 11 Break-up 70 5 10 90

276 12 Break-up 60 5 20 80

301 13 Break-up 50 5 50 50

326 14 Break-up 40 5 60 40

351 15 Break-up 30 5 70 30

376 16 Break-up 25 0 80 20

401 17 Break-up 20 0 90 10

426 18 Park 15 0 98 2

451 19 Park 5 0 98 2 Yes

476 20 Park 5 0 98 2 Yes

LARGE HERBIVORE INFLUENCE ON NATURAL WOODED LANDSCAPE 413

5 Trees are only old enough to be considered break-up stands) (Figure 4; Table 3). Some con-
veteran (>450 years old) towards the end of the tiguous patches of park phase were several
grove and break-up phases, but a few might hundred metres across and at least 5 ha in size in
then survive through the next park phase, a more or less compact block. When the land-
before declining as the scrub and young trees scape was divided into tetrads of 200 m ϫ 200 m
re-established. Any such survivors from the more than 90 per cent contained at least two
previous cycle were assumed to be completely phases and 45 per cent cells three phases. There
lost through shading-out during the first would thus be a high degree of small-scale struc-
50 years into the next grove phase (column 8, tural diversity and ‘edge effects’, e.g. where grove
Table 2). and park cell were adjacent or a scrub cell was
next to a break-up cell.

Results Scenario 2: ‘clumped mosaic’ Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
Clumping the cells in a particular phase did not
Landscape characteristics of different scenarios alter the proportions of the different phases, but
gave a coarser-grained landscape with some open
Scenario 1: ‘wooded landscape’ park patches up to 125 ha (Table 3). The larger
This produced a predominantly wooded, fine- areas of ‘core’ habitat, large contiguous patches
scale mosaic (50 per cent grove, 10 per cent of a single phase, might suit better species with
large minimum patch-size requirements and also

Park Grove
Scrub Break-up

5

Northing (km) 4

3

2

1

0 1 234 5
0
Easting (km)

Figure 4. Distribution of all phases in year 1 using the Woodland mosaic (scenario 1).

414 F O R E S T RY
Table 3: Summary of the composition of landscapes produced by different scenarios

% landscape which is under different phases

Model scenarios Park Scrub Grove Break-up

1. Wooded landscape 25 15 50 10
2. Clumped mosaic 25 15 50 10
3. Cycle length varying across landscape 25 16 47 12
(40, 25, 10) (30, 15, 15) (20, 50, 60) (10, 10, 15)
(value for different sections in west, centre and
east of landscape, see Table 1) 24–26 14–16 49–51 9–11
4. Patchy variation in cycle length across landscape 25 20 20 35
5. Savannah

be more attractive to aurochs, if, like the free- forage from the grassland vegetation would be Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
ranging cattle in the New Forest (Hampshire, greater in the savannah scenario, whereas the
UK), they preferred to graze large areas (>10 ha) wooded-landscape scenario provides more cover
(Putman, 1986). and potential winter browse. Veteran
(>450 years) and near-veterans (301–450-year-
Scenario 3: variable cycles across the landscape old trees) were found across both landscapes (30
Allowing different sections of the landscape to per cent of cells for each group).
operate on different cycles resulted in more open
park (40 per cent) in the west, whereas in the east Both scenarios provided continuity of grass-
the grove dominated (60 per cent). However, the land and woodland-type vegetation over the
proportions of the different phases for the whole landscape and over time, albeit in different
5 ϫ 5 km landscape were little different to those amounts. The interspersion of the different
for earlier scenarios (Table 3; Figure 5). phases (grove next to park, for example) would
allow even relatively poor-colonizing species, e.g.
Scenario 4: variable cycle lengths for individual ancient woodland indicator plants (Brunet and
cells von Oheimb, 1998) to adapt to the shifts in
Over the whole landscape varying the cycles occurrence of open or closed canopy conditions
for individual cells between about 250 and across time. Both scenarios provided scope for
1000 years had little effect on the extent of the birds of scrub or understorey woodland.
different phases across the 500-year period; from
592 to 649 ha for the park phase, and 1219 Discussion
to1280 ha for the grove phase (Table 3).
The role of large herbivores in past landscapes
Scenario 5: savannah has almost certainly been underestimated. Tree
The landscape still had a high tree cover, but regeneration can occur in the presence of large
more of the landscape was in the break-up rather herbivores through the associated protection
than the grove phase (Table 3). provided by thorny scrub as described by Olff et
al. (1999) and Bakker (2003). However, were the
The changing extent of different cover types and herbivore populations high enough to make this
habitat types the commonest type of regeneration? It is difficult
to derive herbivore densities directly from
The Savannah was more open than the wooded- archaeological or other remains (Vera, 2000).
landscape scenario (Table 4), but 55 per cent of Maroo and Yalden (2000) suggest population
the landscape had tree cover greater than 25 per sizes for different herbivore species, but these are
cent, and 35 per cent more than 50 per cent tree based on assumptions about the vegetation com-
cover. The availability of high quality summer position and analogies to species densities in

LARGE HERBIVORE INFLUENCE ON NATURAL WOODED LANDSCAPE 415

Park Grove
Scrub Break-up
5

Northing (km) 4

3 Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
2
1

0 5
01 2 3 4

Easting (km)

Figure 5. Phase distribution for variable cycles across the landscape (scenario 4).

Bialowieza Forest; therefore these figures cannot the parts of the landscape under scenario 3 (from
then be used to infer the nature of the past vege- 40 per cent to 10 per cent park, 20–60 per cent
tation. Models, even simple and mechanistic ones grove) (Table 1). Other evidence must be used to
such as that presented here, help to bridge the suggest which of these outcomes may be the more
gaps in the real observations. likely.

The key assumptions in this model concern the Foraging patterns may favour more grove
cell size and the length of the phases. The model
can be run at any cell size, e.g. if the cells were Cattle tend to concentrate on grassland rather
20 ϫ 20 m they would more or less represent the than in woodland for forage if the former is avail-
space occupied by single mature trees. Cell size able (Wallis de Vries, 1994; Hulbert, 2002; J.
(and whether or not they are clumped as in Bokdan, unpublished). In the New Forest, a
scenario 2) affects the spatial scale of the vari- mosaic of open grassland, heath, mire and closed
ation, but not the overall proportions of the woodland with free-ranging cattle and ponies,
different phases in the landscape. Assumptions 50 per cent of the animal grazing time was spent
about the phase length, and whether this is the on the grassland, even though it occupied only 7
same for all cells, affect how open the landscape per cent of the extent (Putman, 1986).
is: for example, compare the differences between

416 F O R E S T RY

Woodland forage is more likely to be used in 150 year-olds), and that there is then 50 per cent Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
winter when the grassland resource has been mortality of these over each successive 150-year
exhausted and, to some extent, at night when the period. If this pattern is sustained then the
animals move into the wood for shelter. With average density would be 45 trees ha–1, consist-
year-round grazing, winter forage quantity and ing of 24 trees aged 75–150 years, 12 trees aged
quality may be the limiting factor on animal 151–300 years, six trees aged 301–450 years,
numbers. Putman (1986) estimated the natural and three veteran trees (>450 years). The mean
carrying capacity of the New Forest vegetation at inter-tree distance (trunk to trunk) would be
different months of the year for ponies as only only 15 m, whereas the crown diameter for most
about 200–400 between November and January, of the trees would be more than this (Kirby,
but between 2000 and 3000 animals during the 2003); i.e. the canopies would touch over large
summer months. areas. Lower rates of regeneration, or higher
rates of mortality, could produce more open
If available forage is in excess of the summer landscapes, but seem, to this author, less likely.
population needs then marginal areas (in forage
terms) may not be grazed as heavily; more Paleoecological evidence for woodland as the
regeneration of woody species will occur and the matrix rather than more open habitats
area of the preferred forage (grassland) will
decline. In the New Forest the heathland is cut or High levels of tree and shrub pollen tend to be
burned to suppress tree invasion (Putnam, 1986) recorded in profiles covering the Atlantic forest
because there is a tendency for heath to move to maximum in Britain. In modern landscapes
scrub and woodland faster than the woods open values for non-arboreal pollen of <15 per cent
up to form new open heath. Kramer et al. (2003) usually reflect wooded landscapes (<30 per cent
similarly concluded that in Veluwe (The Nether- open land) (Brostrom et al., 1998; F.J. Mitchell,
lands) fire as well as grazing might be needed to personal communication).
prevent the heath turning into woodland.
Perhaps natural fires also played a more signifi- Svenning (2002) compared the interpretations
cant part than has often been assumed in keeping that might be drawn about the landscape for sites
landscapes open in Britain. where the pollen record could be set alongside
other evidence such as beetle and mollusc
Only a relatively small proportion of the land- remains. He concluded that the vegetation for
scape needs to be of relatively high quality forage most areas would be closed, old-growth forest
(such as would be provided by the open park with localized longer-lasting openings around
phase) to support the animals in summer at the ponds and some grassy glades. Open vegetation
population level that can be sustained over the would be present mainly on floodplains, on cal-
whole landscape in winter. The area not being careous (Bush and Flenley, 1987) or poor sandy
used intensively in summer would tend to scrub soils.
up, and hence this suggests that Scenario 1, where
the grove is twice the park phase in extent is a Dirkx (2002) comments that dung beetles had
more likely situation than scenario 5 (savannah) been common before the Atlantic period, then
or the (long park area within scenario 3) where declined during the presumed forest maximum
the park phase is more extensive than the grove. (suggesting fewer large herbivores), before increas-
ing again in the Neolithic period as farming led to
Regeneration pressure may favour grove increases in the numbers of stock (and hence
formation dung). Wood- and tree-dependent Coleoptera were
relatively more common in Mesolithic remains in
Very low densities of regeneration can produce Britain, whereas pasture/dung beetles were rela-
a high percentage canopy cover, even if it is tively more abundant in Neolithic sites (Robinson,
assumed that trees establish as isolated indi- 2000).
viduals. Suppose that, within a regeneration
window of 75 years, just 24 oak trees per This suggests that the model scenarios with
hectare establish to become mature (75– more grove values than open park may be more
likely than the converse.

LARGE HERBIVORE INFLUENCE ON NATURAL WOODED LANDSCAPE 417

Mixed landscape models The relevance of our understanding of natural
landscapes for cultural landscape conservation
In developing the model the whole landscape was in Britain
assumed to go through all the phases of Figure 1,
although Vera (2000) acknowledges that not all The former natural landscape has gone and
areas would be subject to this cycle. In practice cannot be re-created in Britain. The priorities for
there could have been at least three different conservation, e.g. as set out in the UK Biodiver-
landscapes – those permanently open, those per- sity Action Plan (English Nature, 1998), are the
manently woodland (with more shade-tolerant remnants from the last 3000 years of cultural
trees and shrubs), and those where Vera’s cyclical activity, with their associated species. The
vegetation turnover applied. These landscapes conservation of chalk grassland in the form in
may themselves have been in flux over longer which it has existed for much of the last
periods and larger spatial scales (Figure 6) than 1000 years usually involves agricultural-type
have been considered here. grazing regimes, not a reversion to ‘more natural’
grazing systems which might promote scrub
invasion. Bokdam and Gleichman (2000) and Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
Bokdam (2002) similarly found that free-range
grazing did not by itself prevent invasion of
heathland by pine and birch.

Herbivore population
more-or-less stable
Vera cycle

Herbivores Permanent Herbivores
declining open space increasing

Permanent
grove

Low High
Landscape accessibility to herbivores

Figure 6. How differences in space (accessibility to herbivores) or over time (in terms of herbivore
population dynamics) might shift the balance between the basic Vera cycle, permanent open space and
permanent groves over long time-scales and large spatial scales.

418 F O R E S T RY

The model outcomes do, however, suggest acceptability, public safety and animal welfare Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
that, if large herbivores influenced habitat that need to considered before any proposal can
patterns in the ways indicated, species associated be adopted (e.g. Korthals et al., 2002; Klashorst
with open habitats, as well as those of closed and Kreetz, 2002; Kampf, 2002a, b; Limpens et
woodland, could have been common in the past. al., 2002). However, despite the challenges
Similarly, while trees in the centre of groves may involved, there is scope for exploring naturalistic
have grown up as tall, unbranched individuals, grazing regimes as an alternative to traditional
such as have been found in peat deposits farming or forestry practice in some areas
(Rackham, 2003), there could also be trees at the (Worrell et al., 2002; Rogers and Taylor, 2003).
edges between patches (the cells in the model)
that would have developed low branches, and a Acknowledgements
scatter of veterans standing in the open park
phase. This diversity of conditions has been con- My thanks to the colleagues inside and outside
trasted to the landscape usually envisaged under English Nature who contributed ideas and com-
the alternative gap-dynamic approach with its mented on the report from which this paper is taken,
much lower levels of open space. However, in or on the paper itself, including Keith Alexander,
most cases, there is little evidence that open Isabel Alonso, Helen Armstrong, John Bacon, Sallie
ground species (other than some plants) were Bailey, Jan Bokdam, Keith Duff, Rob Fuller, Ted
common in the prehistoric landscape; current Green, Alison Hester, John Hopkins, Loek Kuiters,
abundance of these species cannot therefore be Jim Latham, Fraser Mitchell, John Patmore, George
taken as a sign of what past landscapes were like. Peterken, Patrick Roper, Neil Sanderson, Frans Vera,
Jon Webb and Tony Whitbread.
Vera’s ideas are also relevant to the develop-
ment of low-intensity ways of managing large- References
scale landscapes, provided the unpredictability of
this approach is recognized. Landscapes under a Bakker, E.S. 2003 Herbivores as mediators of their
Vera cycle are not necessarily predominantly environment: the impact of large and small species
open – the model suggests a wide range of con- on vegetation dynamics. Ph.D. thesis, Wageningen
ditions could develop. University, Wageningen.

Current areas of high value grassland and Bokdam, J. 2002 Grazing as a management tool in
heathland areas may be lost to scrub and then naturalised semi-natural heathland in the Nether-
woodland for tens or hundreds of years; even in lands. In Proceedings of the Sixth National Heath-
the long park element (Table 1) there are land Conference J.C. Underhill-Day and D. Liley
250 years between the start of the scrub and end (eds). Royal Society for the Protection of Birds,
of the grove phase. The grazing area must be Sandy, pp. 25–40.
large enough for replacement areas to develop in
an acceptably short time. Bokdam, J. and Gleichman, J.M. 2000 Effects of
grazing by free-ranging cattle on vegetation
Some species may be lost permanently if the dynamics in a continental north-west European
micro-habitats they depend on occur at too low heathland. J. Appl. Ecol. 37, 415–431.
a density or at too infrequent intervals under free-
range grazing. For some species there may not be Bradshaw, R.H.W., Hannon, G.W. and Lister, A.M.
other source populations near enough to allow 2003 A long-term perspective on ungulate-vegeta-
for recolonization when conditions do again tion interactions. For. Ecol. Manage. 181, 267–280.
become suitable.
Brostrom, A., Gaillard, M-J., Ihse, M. and Odgaard, B.
In some conditions the Vera cycle may not 1998 Pollen-landscape relationships in modern
function, e.g. if the area is very fertile and heavily analogues of ancient cultural landscapes in southern
used by animals, or alternatively if it is hardly Sweden – a first step towards quantification of vege-
accessible to them; such landscapes may move tation openness in the past. Veget. Hist. Archaeobot.
towards permanent open habitat or permanent 7, 189–201.
woodland (Figure 6), even under free-range
grazing. Brunet, J. and Oheimb, G. von 1998 Migration of
vascular plants to secondary woodland in southern
There are also issues relating to public Sweden. J. Ecol. 86, 429–438.

Buckland, P.C. 2002 Conservation and the Holocene
record: an invertebrate view from Yorkshire. Bull.
Yorks. Natur. Union (supplement) 37, 23–40.

LARGE HERBIVORE INFLUENCE ON NATURAL WOODED LANDSCAPE 419

Bush, M.B. and Flenley, J.R. 1987 The age of chalk Mellanby, K. 1968 The effects of some mammals and Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016
grassland. Nature 329, 434–436. birds on the regeneration of oak. J. Appl. Ecol. 5,
359–366.
Dirkx, G.H.P. 2002 Livestock farming changes the
landscape. Vakblad Natuurbeheer. Special issue. Ollf, H., Vera, F.W.M., Bokdam, J., Bakker, E.S.,
Grazing and grazing animals, May 2000, 19–21. Gleichman, J.M., De Maeyer, K. and Smit, R. 1999
Shifting mosaics in grazed woodland driven by the
Dobson, A. and Crawley, M. 1994 Pathogens and the alternation of plant facilitation and competition.
structure of plant communities. Trends Ecol. Evolut. Plant Biol. 1, 127–137.
9, 303–398.
Peterken, G.F. 1981 Woodland Conservation and
English Nature 1998 UK Biodiversity Group: Tranche Management. Chapman and Hall, London.
2 Action Plans, Vols 1 and 2. English Nature, Peter-
borough. Peterken, G.F. 1996 Natural Forests. Cambridge
University Press, Cambridge.
Fuller, R.J. 1995 Birdlife of Woodland and Forest.
Cambridge University Press, Cambridge. Peterken, G.F. and Tubbs, C.R. 1965 Woodland
regeneration in the New Forest, Hampshire, since
Green, T. 1992 The forgotten army – woodland fungi. 1650. J. Appl. Ecol. 2, 159–170.
Br. Wildlife 2, 85–86.
Putman, R. 1986 Grazing in Temperate Forest Ecosys-
Green, T. 2002 The role of invisible diversity in tems: Large Herbivores and the Ecology of the New
pasture-landscapes. In Pasture Landscapes and Forest. Croom Helm, London.
Nature Conservation. B. Redecker, P. Finck, W.
Hardtle, U. Riecken and E. Schroder (eds). Springer- Rackham, O. 2003 Ancient Woodland, rev. edn.
Verlag, Berlin, pp. 137–145. Edward Arnold, London.

Hulbert, I.A.R. 2002 Livestock grazing of woodlands – Read, H. 2000 The Veteran Trees Management
impact and management options. Scot. For. 56, 5–17. Handbook. English Nature, Peterborough.

Kampf, H. 2002a Nature conservation in pastoral Robinson, M.A. 2000 Coleopteran evidence for the
landscapes: challenges, chances and constraints. In elm decline, Neolithic activity in woodland clearance
Pasture Landscapes and Nature Conservation. B. and the use of the landscape. In Plants in Neolithic
Redecker, P. Finck, W. Hardtle, U. Riecken and E. Britain and Beyond. A.S. Fairburn (ed.). Oxbow
Schroder (eds). Springer-Verlag, Berlin, pp. 15–30. Press, Oxford, pp. 27–36.

Kampf, H. 2002b Large herbivores and government Rogers, S. and Taylor, K. 2003 New wildwoods:
policy. Vakblad Natuurbeheer. Special issue. Grazing removing barriers to development and implemen-
and grazing animals. May 2000, 56–59. tation. LUPG Report. Joint Nature Conservation
Committee, Peterborough.
Kirby, K.J. 2003 What might a British forest-landscape
driven by large herbivores look like? English Nature Spencer, J.W. 2002 Managing wood pasture landscapes
Research Report No. 530. English Nature, Peter- in England: the New Forest and other more recent
borough. examples. In Pasture Landscapes and Nature
Conservation. B. Redecker, P. Finck, W. Hardtle, U.
Klashorst van den, M.P.M.H. and Kreetz, R. 2002 Riecken and E. Schroder (eds). Springer-Verlag,
Communication and grazing at Natuurmonu- Berlin, pp. 124–136.
menten. Vakblad Natuurbeheer. Special issue.
Grazing and Grazing Animals. May 2000, 49–50. Svenning, J.-C. 2002 A review of natural vegetation
openness in north-western Europe. Biol. Conserv.
Korsthals, M., Keulartz, H., van den Belt, H., Klaver, 104, 133–148.
I. and Gremmen, B. 2002 Battles of nature: the
ethical side of grazing by large herbivores. Vakblad Tubbs, C.R. 1986 The New Forest. Collins, London.
Natuurbeheer. Special issue. Grazing and Grazing Tubbs, C.R. 1996 Comment – wilderness or cultural
Animals. May 2000, 43–45.
landscapes: conflicting conservation philosophies.
Kramer, K., Groen, T.A. and van Wieren, S.E. 2003 The Br. Wildlife 7, 290–296.
interacting effects of ungulates and fire on forest Vera, F.W.M. 2000 Grazing Ecology and Forest
dynamics using the model FORSPACE. For. Ecol. History. CAB International, Wallingford.
Manage. 181, 205–222. Vera, F.W.M. 2002a A park-like landscape rather than
a closed forest. Vakblad Natuurbeheer. Special issue.
Limpens, H., Lejeune, M. and van der Veen, J. 2002 Grazing and grazing animals. May 2000, 13–15.
Urbanised man and the longing for the new wilder- Vera, F.W.M. 2002b The dynamic European forest.
ness. In Pasture Landscapes and Nature Conser- Arboricult. J. 26, 179–211.
vation. B. Redecker, P. Finck, W. Hardtle, U. Riecken Wallis de Vries, M.F. 1994 Foraging in a landscape
and E. Schroder (eds). Springer-Verlag, Berlin, pp. mosaic. Ph.D. thesis, Wageningen University,
313–328 Wageningen.
Whitbread, A.M. 1991 Research on the ecological
Maroo, S. and Yalden, D.W. 2000 The Mesolithic effects on woodland of the 1987 storm. Research
mammal fauna of Great Britain. Mammal Rev. 30,
243–248.

420 F O R E S T RY

and Survey Report No. 1.40. Nature Conservancy K., Knightbridge, R. and Brown, N. 2002 New wild-
Council, Peterborough. woods: developing the role of large-scale new native
Whitehouse, N.J. 2000 Forest fires and insects: woodlands in the uplands. LUPG Report. Joint
palaeoentomological research from a sub-fossil Nature Conservation Committee, Peterborough.
burnt forest. Palaeogeog. Palaeoclimat. Palaeoecol.
164, 231–246. Received 8 September 2003
Worrell, R., Peterken, G.F., Scott, A., Pryor, S., Taylor,

Downloaded from http://forestry.oxfordjournals.org/ by guest on August 21, 2016