|
II. Scientific Evidence Regarding Rainforest Ecology and Protection
6. VICTORIAN RAINFOREST FIRE ECOLOGY
It has long been recognised by foresters, botanists and ecologists
that under appropriate fire regimes in south eastern Australia,
sclerophyll forests would replace rainforests and that, in the
absence of fire, rainforests on many sites would replace adjacent
sclerophyll vegetation (Casson 1952, Gilbert 1959, Cremer 1960,
Cremer and Mount 1965, Webb 1968, Howard 1973a,b, Ashton and Frankenberg
1976, Ashton 1981a, Brown and Podger 1982, Smith and Guyer 1983,
Hill and Read 1984, Ellis 1985, McMahon 1987, Ash 1988). Knowledge
of eucalypt regeneration ecology has led to the extensive use
of high intensity burning of logging residue and broadcast sowing
of seed in logged areas with the purpose of encouraging the regeneration
of eucalypts after clear felling (Frankcombe 1961, Ellis 1971,
Gilbert and Cunningham 1972, Ellis and Pennington 1992, Attiwill
1994b). If the ecotone adjacent to clearfall areas burns, it
will be detrimental to competing rainforest species. Although
it is not the primary motivation, these fires may act to reduce
the likelihood and severity of wildfire, but carry the cost of
the potential for fire damage to rainforest species in the buffer
from the site preparation burn itself (Horne and Hickey 1991).
The moisture differential between eucalypt forest and rainforest
has been used to limit the spread of fire (CFL 1987, 1989), usually
in fuel reduction burns in damp forest containing warm temperate
rainforest. If fire breaks are used instead, the risk to downhill
rainforest vegetation in buffers is slight. Boundary tracks may
have impacts of their own including increased soil disturbance
and runoff, and elevated pioneer and weed invasion (see discussion
below). However, it is not possible to be prescriptive about
the use of boundary tracks or fire breaks without taking into
account the character and local conditions of an area.
There is no doubt that southern Australian cool and warm temperate
rainforests occasionally burn under natural disturbance regimes
(Cameron 1992). For example, Neyland and Brown (1994) found charcoal
in nearly all surface soils collected from 59 cool temperate rainforest
stands in eastern Tasmania. The prevalence of fire in temperate
rainforest communities is reflected in the ecological dynamics
of these communities. Ellis (1985) described successional relationships
among 12 vegetation types from the highlands of north-east Tasmania,
including four postulated pathways by which rainforest may become
established. The time taken for succession to proceed from eucalypt
forest to rainforest depends on the fire history and topographically
determined local climate. The long association between rainforest
and fire may also be reflected in the possibility that fire may
have shaped the evolution of the regeneration characteristics
of many rainforest species. For example, A. smithii is
resistant to irregular burning because it develops a lignotuber
which confers a coppicing ability although the species does not
coppice as successfully as adjacent eucalypts (Johnston and Lacey
1983). Most Victorian temperate rainforest species are not obligate
reseeders and are able to resprout after fire (McMahon 1987, Chesterfield
et al. 1990). McMahon (1992) reported that 39% of the
perennial species of East Gippsland warm temperate rainforest
are obligate reseeders and 41% have the ability to resprout.
One important exception is Pittosporum undulatum, a primary
warm temperate rainforest species and dry rainforest canopy species,
which has no obvious means of resisting fire beyond seed stores
and weak resprouting following crown damage.
The presence or absence of disturbance is not at issue in the
management of Victorian rainforests. Rather, the frequency and
intensity of fire and its spatial and temporal distribution are
central to effective management. Frequent, hot fire will eliminate
rainforest from a site. The importance of these factors is illustrated
by the fact that A. smithii is eliminated by a single
intense fire and by recurrent, frequent fire, even though the
species is resistant to infrequent fire that involves canopy scorch
only. Thus, Ashton and Frankenberg (1976) noted that A. smithii
rainforest on Wilson's Promontory is a climax community only in
fire-sheltered lowland gullies. The importance of fire in determining
the distribution of temperate rainforest was emphasised by the
results of distributional and bioclimatic studies of N. cunninghamii.
Howard and Ashton (1973), Howard (1981), and Busby (1986)
suggested that recurrent fire limits the distribution of N. cunninghamii,
interrupting the expansion of the species to suitable habitat
in the Central Highlands of Victoria. Busby (1986, 1987, 1992)
concluded that Victorian primary cool temperate rainforests occupy
only about 20% of their potential area, largely because of fire
frequency. Thus, when considering rainforest protection, one
must consider the impacts of human activities on fire frequency
and intensity.
Cameron (1992) suggested that fires normally function to stabilise
and maintain sharp boundaries between rainforest and sclerophyll
vegetation, 'except under conditions of exceptional fire hazard
such as prevailed in the 1982/83 fire season' (Cameron 1992, p. 20).
In more usual conditions, the steep moisture gradients between
the two communities ensure that the likelihood of spread of fire
into rainforest is less than the likelihood of spread into adjacent
sclerophyll communities. Similarly, Greenlef (1990) noted that
moisture gradients are important determinants of vegetation patterns
in each of five different historical fire regimes in Sequoia
sempervirens communities in the western United States.
Within a fire interval, there is ample evidence that Victorian
rainforest species will tend to spread into adjacent sclerophyllous
vegetation, contracting abruptly with the ensuing fire event.
Cameron suggested that if the fires are sufficiently intense
and frequent, then sclerophyll vegetation will dominate and even
permanently replace, rainforest (see also McMahon 1987). The
long term prognosis for rainforest stands versus sclerophyll forest,
depends on the frequency, intensity and variability of fire events.
If the fire interval is short, or there is a high probability
of successive fires with a relatively short fire interval, then
it is likely that rainforest will be lost. The likelihood that
rainforest will be lost after a fire is a function of the intensity
of the fire, and of the post-fire regenerative success of eucalypts.
Cameron (1992 p. 22) postulated 'moderately successful' initial
recovery following a single fire catastrophe, and progressive
structural collapse and ultimate elimination of rainforest following
'at most two further fire events'. That is, he suggested that
secondary succession may be truncated by successive fire events
if the fire interval is short. The distribution of rainforest
and rainforest ecotones reflects both the detailed history of
a site, in particular its fire history, and the underlying environmental
gradients, especially on sites where gradients are gentle, and
ecotones are broad and diffuse (Melick and Ashton 1991).
Fuel reduction burning and seed-bed preparation result in an element
of risk of the burn escaping to adjacent rainforest. Current
rotations at one coupe mean exposure to such risk every 60-100
years for some stands. Cameron (1992) speculated that, in addition,
adjacent old sclerophyll forest affords more protection to rainforest
from wildfire than does regrowth forest. This suggestion is based
on three postulates. First, during the juvenile and mature growth
phases, regrowth forest uses more water than old-growth forest
and therefore the soil has lower moisture content. Second, more
mesic species with broad-leaved canopies present in the old-growth
forest will be replaced by more flammable, sclerophyll understorey
species. Third, Cameron speculated that the greater height of
the canopy fuel in old-growth forest contributes to a fire shadow
in circumstances when crown fire may threaten rainforest.
Old forests occupy the least fire prone sites in the landscape.
Topography, moisture levels, the composition of the understorey
and the height of the canopy fuel may combine to reduce fire intensity
and the severity of crown damage. In general, fire intensity
is determined by the amount of fuel, its dryness and to some extent
its vertical spatial distribution. Fuel loads in ash forests
typically are very high in any forest type at the pole stage and
beyond.
There is sufficient empirical evidence to suggest that some of
Cameron's (1992) concerns have scientific justification. McMahon
(1987, 1992) evaluated the impact of the 1982/83 wild fires on
Victorian rainforest stands and summarised his qualitative observations
as follows. Topographical features were one important factor
determining fire intensity at a particular site. Warm temperate
rainforest margins with mixed open canopies of A. smithii
and eucalypt species invariably experienced a crown fire, whereas
closed rainforest experienced surface fires or less intense crown
fires. The fire effects on these two structural types were likely
to be different because of moisture retention and flammability.
The fires stimulated opportunistic species and fire pioneers
usually rare in mature rainforest. Surface fires under a closed
canopy had minimal structural or floristic effects. Narrow stands
in close proximity to eucalypt forest were more susceptible to
crown fire. In areas of complete rainforest canopy removal, post-fire
opportunists dominated the stand. The distribution of sclerophyllous
recruits was related to the degree of closure of the pre-fire
canopy, the level of fire-induced canopy damage, and the availability
of seed in the canopies of emergent eucalypts. Increasing severity
of fire retards the rate of rainforest canopy recovery because
Acmena smithii tends to regenerate from epicormic shoots
and basal resprouts with increasing severity (see also, Chesterfield
et al. 1990). It will only regenerate successfully
from seed in undamaged stands. Abundant recruitment of sclerophyll
species suggested that floristic and structural change following
intense fire may be long-term (of the order of 300-500 years,
the life expectancy of each eucalypt cohort), even in some areas
that were previously closed mature rainforest (McMahon 1992; Woodgate
et al. 1994).
Other studies provide confirmatory, if indirect, evidence that
Cameron's (1992) speculations summarised above have substance.
Brown and Podger (1982) found that the persistence of populations
of Huon pine, King Billy pine, pencil pine (Athrotaxis cupressoides)
and deciduous beech (Nothofagus gunnii) in cool temperate
Tasmanian rainforest depend on the absence of fire for prolonged
periods. Hill (1982) and Hill and Read (1984) found fire behaviour
in cool temperate rainforest to be an important determinant of
post-fire regeneration. Fire intensity and proximity of sclerophyll
seed source may lead to long term structural and compositional
change. Barker (1991) documented a population of the native conifer
Podocarpus ssp. nov. (Ross 1993) in cool temperate
rainforest at Goonmirk Rocks in Victoria. Barker suggested that
its persistence in this unique habitat is facilitated by the broken
canopy of the marginal habitat and the long-term absence of fire.
Chesterfield et al. (1991) noted that populations of E. nitens
which have since been described as the Victorian endemic E. denticulata,
have a restricted distribution, even though the species is a relatively
vigorously growing eucalypt. They suggested that it competes
for habitat with rainforest understorey species in locations where
fire is rare, and that it is relatively shade intolerant.
Overall, there are insufficient data available to judge conclusively
the validity or otherwise of Cameron's (1992) arguments, and they
remain speculative. The interaction between fire frequency, regeneration
niche, and competitive tolerance is a pervasive feature of the
ecology of Victorian forests. For example, it is responsible
for structuring the boundaries between E. obliqua
and E. regnans (Ashton 1981). Generally, the ignition
probability and intensity of fires in these forests are closely
related to fuel moisture content. Thus, our ability to predict
accurately the impacts of management on the distribution of rainforest
in the medium to long term will depend on our ability to understand
the relationship between topography, climate, fuel loads, fuel
moisture content, fire intensities, fire probabilities and other
stochastic processes, and the competitive interactions among species.
Such understanding is an essential component of strategic planning
for rainforest protection.
6.1 Characteristics of natural fires
Given that natural disturbance regimes are the appropriate model
for forest management (Attiwill, 1994b) and that fire is the dominant
ecological process driving disturbance dynamics in Australian
forests (Attiwill 1994a), it would seem important to characterise
the patterns of natural fire events in the past. Characterisation
should include knowledge of fire probability and intensity in
different forest types, and the spatial and temporal pattern of
fire.
Perhaps surprisingly, there is very little known about historical,
natural fire regimes in Victoria, or in other, analogous landscapes
in Australia. King (1963, in Bowman and Jackson 1981) suggested
that fires in high rainfall areas became less frequent but more
intense as a result of colonisation of the Australian environment
by Europeans and elimination of Aboriginal influence on the landscape.
Cremer (1960) noted that a reduction in fire frequencies would
have resulted in the expansion of rainforest species into adjacent
sclerophyll forest. Jackson (1968) suggested that in Tasmanian
forests, areas with fires at 150-200 year intervals carry mixed
forest (rainforest with emergent eucalypts), and areas with lower
frequency fires carry pure rainforest. All of these statements
were essentially speculative, based on little or no direct field
evidence.
Most fires in ash forests occur during relatively dry periods.
Given the amount of fuel and the flammability of the vegetation,
they tend to be major fires. Attiwill (1994b) noted six such
fire events in the last 200 years in south eastern Australia,
occurring in 1851, 1898, 1926, 1932, 1939 and 1983. Vines (1974)
stated that 'very bad' fires occurred in Victoria in 1851, 1886,
1913, 1926, 1939, 1952 and 1965, and that minor forest-fires occurred
in 1920, 1932, 1944 and 1960. Complete fire record dates are
available only from 1910. Vines (1974) suggested that major wildfires
occur about every 13 years in Victorian forests, based on an analysis
of weather patterns and associated fire events. His prediction
was not corroborated by subsequent fires. Such analyses treat
State-wide or regional data and are somewhat irrelevant for ecological
purposes unless quantified for different climate and landscape
zones, and for a particular point in the landscape.
Woodgate et al. (1994) recorded that fire frequencies from
eight trees ranging from regrowth to senescent in an E. sieberi
stand in East Gippsland averaged about every 22 years between
1800 and 1992, and every 30 to 40 years prior to 1800. They concluded
that fires at 20 to 50 year intervals are likely to represent
the long-term average in this ridge community. In the absence
of historical fire data such as these (scant though they are),
any attempt to use natural disturbance regimes as a model for
forest management will be hampered by a lack of necessary information.
It is possible to suggest, for example, that fire frequencies
and intensities in damp forest on ridge tops are very likely
to be different to those in rainforest occurring at the bottom
of gullies or adjacent to wet or montane forest. However, we
do not know the expected frequency or the spatial or temporal
variance of these fires. Such data are essential for specifying
acceptable bounds on forest disturbance in managed landscapes.
Even if the characteristics of natural fire regimes were known,
the silvicultural practices now in operation in the forests surrounding
rainforest are unlikely to closely mimic natural disturbance.
The effects of natural fires are unlikely to be the same as the
effects of logging followed by site preparation burns because
the spatial pattern, intensity, the effects on eucalypt, tree
fern and understorey tree survival, and the distribution of surviving
individuals as well as the distribution and type of fuels that
remain after these disturbances will be different (eg., Mount
1979, Mueck and Peacock 1992, Ough and Ross 1992, Kutiel and Shaviv
1992; see the example in Figure 3). When State Forests available
for timber production in Victoria burn, they are usually salvage
logged. In contrast, natural fires leave behind much greater
structural complexity in the form of dead stags, logs, recovering
trees, resprouting shrubs and ferns than is normally found in
areas that are clear felled, or areas that are salvage logged
after natural fire (Franklin 1988, 1992, 1993). Furthermore,
regeneration burns are likely to be smaller in area than other
fires. For example, the fires of 1939 burned a larger area than
will be logged and regenerated in the following 80 years. Wildfires
may create more complex edges than are found in logged landscapes
(Franklin 1993). They are likely to burn areas that would not
normally be logged such as stream sides and forests on steep terrain.
These differences are likely to influence many factors including
wind dynamics and fuel loads, and hence future fire probabilities
and intensities.
If we ignore, for the moment, the spatial extent of fires, we
may examine the effect of random intervals between fires on the
expected age structure of a forest when the extent of disturbance
from logging and natural fire regimes is the same (Fig. 3).
The deterministic regime is represented by a 200 year rotation,
and one block is harvested each year. The fire regime is equivalent
to a geometric random variable. Consider a forest as being composed
of numerous areas. If each area is disturbed with an annual probability
of p, then the expected proportion of areas that are n
years old is equal to the proportion of areas disturbed n
years ago (=p) multiplied by the probability that these
areas are not disturbed subsequently (=[1-pn-1).
Therefore, the age structure of the forest is described by the
function which gives the proportion of forest that is expected
to be n years old. The average time between disturbance
events is equal to 1/p (after McCarthy and Burgman, in
press).
McCarthy and Burgman (in press) suggested that the result in Figure
3 relates only to the "expected" age composition of
the forest, representing the mean structure. There is no indication
of fluctuations in age structure over time. If disturbance occurs
at a small scale relative to the forest, such as may be due to
individual treefall, or harvesting a small proportion of the forest
each year, then the age structure of the entire forest will not
fluctuate very much. At a finer scale (ie., at the scale of disturbance) |
