FOREST DYNAMICS MODELLING UNDER NATURAL FIRE CYCLES: A TOOL TO DEFINE NATURAL MOSAIC DIVERSITY F O R FOREST MANAGEMENT SYLVIE G A U T H I E R Canadian Forest Service, Petawawa National Forestry Institute, PO. Box 2000, Chalk River, Ontario, KOJ 1JO, Canada and A L A I N L E D U C and YVES B E R G E R O N Groupe de Recherche en Ecologie Forestidre ( GREF), Universit# du Qudbec ,~ Montreal, C.P. 8888, Succ. Centre-~lle, Montreal, Qudbec, H3C 3P8, Canada

Abstract. In natural boreal forests, disturbances such as fire and variation in surficial deposits create a mosaic of forest stands with different species composition and age. At the landscape level, this variety of stands can be considered as the natural mosaic diversity. In this paper, we describe a model that can be used to estimate the natural diversity level of landscapes. We sampled 624 stands for tree species composition and surficial deposits in eight stand-age classes corresponding to eight fire episodes in the region of Lake Duparquet, Abitibi, Qu6bec at the southern fringe of the Boreal Forest. For six surficial deposit types, stand composition data were used to define equations for vegetation changes with time for a chronosequence of 230 years for four forest types. Using Van Wagner's (1978) model of age class distribution of stands, the proportion of each forest type for several lengths of fire cycle were defined. Finally, for real landscapes (ecological districts) of the ecological region of the "Basses-Terres d'Amos", the proportion of forest types were weighted by the proportion of each surficial deposit type using ecological map information. Examples of the possible uses of the model for management purposes, such as biodiversity conservation and comparisons of different landscapes in terms of diversity and sensitivity to fire regime changes, are discussed.

1. Introduction A current priority in forest management is the development of practices aimed at forest ecosystem sustainability. Factors other than timber production, such as recreation, wildlife, and biodiversity conservation must also be considered in the context of sustainable forestry. The last factor has gained much prominence in the last few years. Biodiversity can be measured at different levels: genetic, species, ecosystem, and landscape (Canadian Forest Service, 1993). However, a speciesby-species approach for measuring biodiversity is difficult because: 1) there are numerous species of which many are unknown, 2) the task would be extremely time consuming, 3) the financial resources are not available, and 4) it will exhaust the general public's patience (Franklin, 1993). Hence, a large scale approach "is the only way to conserve the overwhelming mass of existing biodiversity" (Franklin, 1993; see also Noss, 1983, 1987; Mladenoff et al., 1993; Attiwill, 1994a). A landscape approach can help define a framework to design and manage landscapes for Environmental Monitoring and Assessment 39: 417-434, 1996. (~) 1996 Kluwer Academic Publishers. Printed in the Netherlands.

418

SYLVIE GAUTHIER ET AL.

biodiversity conservation. This is based on the assumption that forest management strategies focusing on the conservation of the diversity of patches in forested landscapes is a way to ensure the maintenance of a biodiversity level similar to that under natural conditions (Franklin, 1993; Noss, 1987). In natural forested landscapes a diversity of forest patches occurs mainly as a result of three sets of components: 1) abiotic conditions (geology, climate, parent material, etc.), 2) disturbance events (fire, insects, diseases, etc.), and 3) chance events (dispersal, pre-disturbance forest composition, etc.). In the boreal forest, among abiotic factors the surficial material type is important in accounting for differences in vegetation composition and dynamics from site to site (Bergeron et al., 1983; Bergeron and Bouchard, 1984; Sims et al., 1989). In these forests, fires are the main disturbance type responsible for the dynamics and maintenance of the habitat mosaic (Heinselman, 1981; Attiwill, 1994a). The occurrence of fire creates forest patches that vary in age and composition. The natural fire cycle of the boreal biome is estimated to be approximately 100-150 years (Bergeron, 1991; Johnson, 1992). Given a constant fire frequency and an equal probability of burning for any stand age, Van Wagner (1978) showed that the expected age-class distribution of stands should follow a negative exponential:

F(t) = lib 9 e x p ( - t / b ) , where F(t) is the proportion of stands of a given age, t is time, and b is the fire cycle. A sufficiently large landscape under a constant fire cycle would then present a negative exponential age-class distribution, where 63.2% of the area would be composed of stands younger than the fire cycle length. With this constant age class distribution, the landscape forest vegetation would be in quasi-equilibrium with the fire cycle (Shugart, 1984). Our paper presents a simple model that allows the estimation of mosaic diversity of forest landscapes in quasi-equilibrium with a particular fire cycle. The model integrates two sets of components, disturbance events and abiotic factors, to evaluate the diversity of landscape mosaics. It was developed for one ecological region near Lake Abitibi, Qu6bec. The model was applied to landscapes that differed in their topography and surficial material composition for comparison purposes. In this paper, the model is described and some examples of its potential use are shown. Finally, limitations and further improvements of the model are discussed.

2. Study Area The ecological region (homogeneous regional climate) of Les Basses-Terres d'Amos covers approximately 16 000 km 2 in the boreal forest of Qu6bec (Table I; Figure 1). Mature forests are characterized by balsam fir (Abies balsamea (L.) Mill.) and white birch (Betula papyrifera Marsh.) (Rowe, 1972; Bergeron and Bouchard, 1984). The region is located in a large physiographic unit, the Clay Belt

2736

Cote~ux du lae joanns

ai de a Rlvr~ren~arrl~ana

B20

309

15 758

Co.t.eau~.de la. rtvlere Magusl

Basses-Terres d'Amos

L049

602

Plaljne de la rlvltre Promenade

362

L040

L026

248

P l a n e d e lac ~urlmau

C22

103

C ~ ~k,leSm~oU/~c

D06

Area (kin ~)

Name

No.

Ecological District

7

13

2

0

4

4

50

Rock

I1

14

9

0

23

59

g

Till

6

2

9

0

40

1

0

Sand

53

48

14

88

14

23

29

Clay

16

11

63

8

14

9

5

Organic

Deposits (percentage age surface cover)

7

12

0

3

3

2

6

Water

0

0

3

0

2

0

2

Other

43

10

14

80

29

31

altitude (m)

Mean .. a~plitude o.f

TABLE I Landscape characteristics of selected ecological districts of Les Basses-Terres d'Amos, including total area, percentage cover by surficial deposits, and the range of altitude.

"~

.< m

0

7

7

r"

~3

~r

Z

~n

420

SYLVIE GAUTH1ER ET AL.

49015 '

#

o

. ~

~

.r .*.~Lac ~ ' ~ Cl

9

, t'~,..-.

%.J

O

I

-

~

~'~;'--...%'-., .,.,!. ."

s r~,

~

""'," " " N .....

./

.....

~.~,~b,

-

N

;

~.~.., ~ -~. -

r/h:::t~.l *

D

:

,"~

,,, " . . - , . . . . .

~~....-~..'..~",,,.

..

t

9 ~"

"'.

........ ;::.

........

....

ua~

".. Amos,:

",,...~.,:,,.:~

~, ,

e

,.

,........

.............

,

L 40

,

,, I~. . . . .

.N~

.-"

~# ' ,

9

aA,~

,-'t~r,~,.,~--,,

P--.~,

1

"....

",,

..,~t~L~,.

"'"" ~ P . . ~ , , 2-.. ,..,.~.::..,., ~ / ~ , :',...... ~ ",.

9

:--:

'...J

". . . . .

ar ~"e t l

.acDuparquet~WI~V"

-, ~-.

"'" i.

}

I. qt~l/,

':".

f

:

,.

"'~

,---,':~--.,.:

.ao,y.,

""

! 20~u.~'.

............

79000 '

Legend ........

m

District limits Ecological region limits Selectedecological districts Lakes

Fig. 1. Location map o f a section o f Los Basses-Terres d ' A m o s , Abitibi, Qu6bec, showing the selected ecological districts.

FOREST DYNAMICS, FIRE CYCLES, AND NATURAL MOSAIC DIVERSITY

421

79*15'

948*30'

79115'

Legend O

Transects Lake Fire limits Fig. 2. Location and fire history map of the vegetation survey area.

of Ontario and Qu6bec, left by post-glacial Lake Badow-Ojibway (Vincent and Hardy, 1977). The region is divided into 42 districts characterized by a recurrent pattern of topography and a distinct proportion of each surficial deposit (Robitaille, 1989). This definition of districts is similar to the definition of a "landscape" by Forman and Godron (1986). Consequently, to avoid confusion, throughout the paper the ecological districts will be referred as landscapes. A subsection of the ecological region, near Lake Duparquet (Figure 2), was selected for a forest vegetation survey. This survey area was selected because its fire history was known (Bergeron, 1991; Dansereau and Bergeron, 1993) and it was a typical example of the districts of the ecological region (Table I). Despite the fact that portions of the ecological region were clear-cut, the survey sector had not been logged, allowing for the reconstruction of natural post-fire succession. Eight major fire episodes have affected the area in the last 250 years (Figure 2; Dansereau and Bergeron, 1993). By establishing numerous transects in each of the

422

SYLVIEGAUTHIERETAL.

eight fire-affected areas, a large distribution of forest ages and types of surficial material was sampled.

3. Methods 3.1. VEGETATIONSAMPLING To define the forest dynamics with time since fire, the forest vegetation (tree species) was sampled in 624 quadrats (16 m x 16 m) on 37 transects dispersed in the area affected by each of eight fire events (Figure 2). Transects were subjectively dispersed to maximize the sample representation of each surficial deposit and minimize transportation among the sample quadrats. In each quadrat, the DBH of all trees > 5 cm was noted by species. The surficial material, drainage, and elevation were also determined. Four forest types were defined based on the relative dominance (basal area) of coniferous or deciduous species: coniferous (> 75% coniferous species), mixed-coniferous (50% < coniferous species < 75%), mixeddeciduous (25% < coniferous species < 50%), and deciduous (< 25% coniferous species). Six site types were also recognized based on surficial deposit: 1) rock (rock and shallow till), 2) till, 3) sand deposit, 4) mesic clay, 5) hydric clay, and 6) organic. Each quadrat was classified within one forest type and one surficial deposit. The relative frequency of each forest type was calculated by surficial deposit for the eight mosaic ages. 3.2. LANDSCAPESSELECTIONAND DESCRIPTION Ecological maps (Minist6re des Ressources naturelles du Qu6bec) provided the following characteristics of each landscape: 1) landscape area, 2) proportion of water, 3) proportion of surficial deposit types, and 4) topographic description (Table I). Six different landscapes, varying in proportion of surficial deposits and topography roughness, were selected for comparison purposes (Table I). The L49 landscape, surveyed for forest cover composition, is a typical example of the ecological region of Les Basses-Terres d'Amos in terms of surficial deposit distribution (Table I). The other five landscapes differed in that each had one dominating surficial deposit type. Clays with mesic and hydric moisture regime were not distinguished in the ecological maps. Because vegetation dynamics differ between these two types of clay (Figure 3), a proportion of 10% of the clay was assigned as hydric clay, based on the proportion of hydric clay observed in a study of two districts in the region (B61and, 1990). The rolling topography of the selected landscapes, typical for the region, made this assumption realistic but the 10% estimate for hydric clay may be an underestimate for flat landscapes (L26, clay-dominated and L40, organic-dominated).

423

FOREST DYNAMICS, FIRE CYCLES, AND NATURAL MOSAIC DIVERSITY

80.

,o /

Rock deposit

80

/ /

70.

\'rill deposit

60.

6O

50,

5O

~.g31 .724

\

/

\,

40 ~-

30.

30

./\

20 84 10

/

\ \

,

\ 2~5 50 80

Me.sic clay

7O

\~7

75

,/'/,,/\\ /~,

- ~o.

\

, ,,/

10 25

100 125 150 175 200 225

50

\,\

\

75 100 125 150 175 200 225

Hydtlc clay

.~s2

'7

70

60

~

5o

/'""Y V

" ,//\\,/'\'\,

o_ 30

/

2o

,,/\\

/\ /\-~

10

ii ,o

\

\\/-'~

'-'~,\

.r3o

9

20 1o

25

50

75

10o 125 150 175 200 225

Time since fire

0

25

50

75

100 125 150 175 200 225 fire

Time since

Deciduous Mlxed-dodduous

Mixed-conlferous Coniferous Fig. 3,

F o r e s t type s u c c e s s i o n for four surficial depos i t types. R 2 v a l u e s o f p o l y n o m i a l r e g r e s s i o n s

are associated with each forest type curve.

3.3.

MODEL DESCRIPTION

The model developed is empirical and deterministic, being based on actual vegetation data from which the dynamics of forest types were deduced. It is also spatially aggregated, i.e., the spatial organization of the patches was not taken into account in modelling. The model is composed of three main modules: 1) vegeta-

424

SYLVIE GAUTHIER ET AL.

tion dynamics/surficial material; 2) theoretical age-class distribution related to fire cycles; and, 3) computation for real landscapes (districts) using the proportion of surficial material deposits provided by ecological maps, here those of the MinistSre des Ressources naturelles du Qu6bec. 3.3.1. Vegetation dynamics/surficial deposit module The relative frequencies of forest types were used in polynomial regression to model the dynamics of the forest types with time since fire for each surficial deposit (Figure 3). Due to an insufficient number of quadrats located on sand and organic deposits, the modelling of the forest dynamics in the two deposit types was based on the literature. For the sand deposit type, we assumed that jack pine was the dominant species in stands younger than 100-150 years, being replaced after that time period by other conifers (Carleton and Maycock, 1978; Cogbill, 1985) and on organic deposits, black spruce was assumed to always be present (Bergeron et al., 1983; Viereck, 1983; Cogbill, 1985). Thus, for both types of surficial deposit, assumptions were made that the coniferous forest type would be maintained through time. 3.3.2. Theoretical age-class distribution related to fire cycles module The negative exponential (Van Wagner, 1978) was then used to estimate the ageclass distribution of the forest stands with a fire cycle of 100 years. An estimate of the natural fire cycle of this length for the region was obtained (Bergeron, 1991; Dansereau and Bergeron, 1993) and is of the order suggested for the Boreal Forest (Turner and Romme, 1994). By changing the fire cycle length from 50 years to 500 years, an evaluation of the effect of fire cycle changes (due to fire suppression or climate change) on the mosaic diversity of the landscape can be obtained (Bergeron and Dansereau, 1993). To estimate the relative frequency of any specific age-class of forest types, the age-class distribution was combined with the polynomial regression results by surficial deposit (Figure 4). 3.3.3. Computation for real landscapes The results of vegetation dynamics, combined with the expected age-class distribution under a fire cycle, were weighted by the real proportions of surficial deposit of each landscape. Thus, an image of the forest cover composition of the landscapes was obtained. The results can be presented in the form of expected age-class distributions (Figure 5), or they can produce a global image of the landscapes in terms of proportion of each forest type (Figure 6). Two patch diversity indices were computed for each landscape using the Shannon diversity index (Shannon, 1948):

H -- - ~ P i ln(pi)

In(c) For age-patch diversity, Pi represents the proportion of a particular 10-year ageclass for each forest type and c = 100 (25 age-classes * 4 forest types). This

i

l

l

,

, coniferous

i

M

I

I

50 100 150 200 250

Age-class

1 0

n! 50 100 150 200 250

;ig. 4. Comparisons of the age-class proportion of the four forest types and age-patch diversity in the selected landscapes under Lfire cycle of 100 years.

L

s

-~ t~




/3

")

5 4 3 2 1 0

0.861

9 /

Deciduous

Typical (L49)

0

I

6

50 100 150 200 250

I

5 4 3 2 1 0 }

I

50 1 0 0 1 5 0 2 0 0 2 5 0

0

Mixed-deciduous

)rganic dominated (L40)

~-

I

0

6

9 8

0

Mixed-coniferous

50 100 150 200 250

0

5

0

i

>

-]

t~

7

,

I

1 o

I

l

1 i

3 2

3 2 0

5

4

5

4

Sand dominated (B20)

r

0.824

Clay dominated (L26)

,

50 100 150 200 250

i

9

10

8

0.857

Till dominated (C22)

8

9 ~

10

9 8

0

0

0 i

2

9 8 7 6 5 4 3

Rock dominated (D06)

O r-"

~.

o~

=

o

10

426

SYLVIE GAUTHIER ET AL.

I

I Coniferous Mixed-coniferous

80 -

0.550

Mixed-deciduous 70 Deciduous

0.752

~ 50 0.959

"5 40

g

m

~ 2o

Q.

10 0

j

t

D6 Rock

C22 Till

i B20 Sand

I L26 Clay

I

I

L40 Organic

Typical

L49

LANDSCAPE

Fig. 5. Comparisonsof the global proportion of the four forest types and patch diversity in the selectedlandscapes. index, by combining age and patch composition, reflects the structural complexity of stands that differ within forest types through time. For patch diversity, pi is the total proportion of each forest type and c = 4.

4. Results and Interpretation The results of polynomial regressions confirmed that forest type dynamics through time differ from one surficial deposit to another (Figure 3; Bergeron and Bouchard, 1984; Bergeron et al., 1983). For all deposit types, deciduous forests, dominated by pioneer species (white birch or trembling aspen), show a decreasing importance > 100 years after a fire, except in mesic clay where trembling aspen forests were present for 150-200 years. The mixed-deciduous forest types showed an increasing proportion in young stands until 100-150 years after a fire, with the exception of the hydric clay deposit where this increase was delayed after 100 years. With the exception of the rock deposit, the mixed-coniferous forest type had its highest frequency after 100-150 years of stand development. The frequency of the coniferous forest type increases with time since fire, reaching a maximum

427

FOREST DYNAMICS,FIRE CYCLES, AND NATURALMOSAICDIVERSITY Rock dominated (D06)

Till dominated (C22)

Sand dominated

(1320)

100 90

1.0 0.9

,~ 80

0.8

.7o

8 60

~

~ 5o

:

0.6 0.7 "i

~

0.5 0.4 0.3

40 ~

3O

~, 2o

0.2

~" 10 0

0.1

j 0 100 200 300 400 5000 100 200 300 400 5000 Coniferous 9 Mixed-coniferous 9 Mixed-deciduous 9

9

Clay d o m i n a t e d

(L26)

Organic dominated (L40

I

~

,

~

I

0.0

100 200 3OO 4O0 5OO Deciduous 9 Patch diversity Typical (IA9) 1.0 0.9

,0o o

90

~

70

0.7 L~

o

50

~ o

4o

0.5 ~ 0.4 o 0.3 ~ 0.2 0.1 0.0

0.8

0

3O

. . . .

lO 0 0

100 200 300 400 5000

i i i i i 100 200 300 400 5000 Fire cycle length

.

6

100 200 300 400 500

Fig. 6. stimated effect of fire cycle changes on the proportion of forest types and patch diversity in the selected landscapes.

after 230 years. After that time, it dominates each surficial deposit type except till, where an equal proportion of mixed-coniferous and coniferous forest was observed (Figure 3). To obtain information about the age-class distribution of a stand in a landscape in quasi-equilibrium with a particular fire cycle, the negative exponential was combined with polynomial regression. The expected proportion of forest types in landscapes was estimated by combining the three modules (the vegetation dynamics, the age-class distribution under a fire cycle of 100 years, and the proportion of surficial deposit in a landscape). These results showed differences among landscapes in terms of forest type composition and diversity due to the differences in dynamics among surficial deposit and the different proportions of deposit types among landscapes (Figure 4). With the exception of the sand-dominated landscape (B20) and the organic-dominated landscape (L40), the young age-classes of the landscapes were dominated by deciduous forests, but this type of forest is almost absent in the intermediate age-classes (100--150 years). The importance of the mixed-forest type varies from one landscape to another. In all landscapes, except

428

SYLVIE GAUTHIER ET AL.

the sand-dominated landscape and the organic-dominated landscape, only a small proportion of each age-class was represented by coniferous forest types. Under a fire cycle of 100 years, 38% of the typical landscape (L49) would be dominated by deciduous forests and 31% by coniferous forest (Figure 5). In the rock-dominated landscape (D06), equal proportions of coniferous and deciduous forest types were observed (35 and 34%, respectively). The clay-dominated landscape (L26) and the till-dominated landscape (C22) showed higher proportions of deciduous forests and lower proportions of coniferous types (41--43% and 2123%, respectively). The sand- or organic-dominated landscapes would mostly be covered by coniferous forests (62 and 76%, respectively). This result could be an artifact of the different modelling procedure but, as suggested in the literature, the forest cover in these two deposit types is likely to remain coniferous through time (Carleton and Maycock, 1978; Viereck, 1983; Cogbill, 1985). The differences in vegetation dynamics among different surficial deposits are important in explaining the mosaic diversity of a landscape. Our results indicated that, even when different landscapes are affected by the same fire cycle length, forest cover composition differs among landscapes due to different assemblages of abiotic conditions. The age-patch diversity index was highest for the rockdominated landscape followed by the typical landscape and the till-dominated landscape (Figure 4). The sand- and clay-dominated landscapes had similar index levels, whereas the organic-dominated landscape had the lowest age-patch diversity index. In terms of patch diversity, the typical landscape, the till landscape, the rock landscape, and the clay-dominated landscape showed similar values (Figure 5). However, the sand- and organic-dominated landscape showed lower patch diversity indices (Figure 5). To evaluate the effect of a change in the length of fire cycles, the six selected landscapes were compared under fire cycles varying between 50 and 500 years (Figure 6). Proportions of coniferous and mixed-coniferous forests increased in all the landscapes with increased fire cycle length, while the proportion of deciduous forest declined. The sensitivity of the landscapes to changes in fire cycle length varied (Figure 6). For instance, the till-dominated landscape showed a slower rate of change than the other landscapes. This different sensitivity of landscapes is also reflected in the patch diversity index. The till-dominated landscape appears to be the least sensitive to a change in fire cycle: a slight increase in patch diversity under fire cycles of 150 to 300 years was observed, followed by a slow decrease when fire cycles were longer. Even under a fire cycle of 500 years (which could represent fire suppression) the patch diversity value was only 5% lower than that with a fire cycle of 100 years. The typical district showed relatively constant patch-diversity indices for fire cycles of 100-150 years, while the diversity indices decreased with an increase in fire cycle length. A decrease of ~ 18% as compared to the baseline, defined with a fire cycle of 100 years, would be observed with fire suppression (fire cycle = 500 years).

FOREST DYNAMICS, FIRE CYCLES, AND NATURAL MOSAIC DIVERSITY

429

5. Discussion

5.1. APPLICATIONSIN FORESTMANAGEMENT In forestry it is now recognized that biodiversity conservation and the maintenance of structure, function, and integrity of ecosystems are necessary in order to reach sustainability (Noss, 1983, 1987; Canadian Forest Service, 1993; Franklin, 1993). As ecosystems are interconnected entities, there is a need to achieve management at the landscape level in order to maintain overall diversity and health (Noss, 1983, 1987; Franklin, 1993). It is also recognized that, to maintain diversity, we should establish a good network of habitat preserves and also, perhaps more importantly, pay increased attention to management of the landscape matrix where habitats are semi-natural and where most of the diversity may be represented (Hansen et al., 1991; Pimentel et aL, 1992; Franklin, 1993). It is also recognized that disturbances are important in maintaining patch diversity and, as a corollary, the implicit species and genetic diversity in landscapes (Pickett and White, 1985). Fire, for instance, with its patchy behaviour in time and space, ensures the simultaneous presence of different seral forest stages on a regional scale (Attiwill, 1994a, 1994b). Thus, the multi-purpose management of forested landscapes should be placed into the ecological framework of natural disturbances. Tools to evaluate and, consequently, to maintain natural patch diversity are needed to effectively manage our forest landscapes. In this paper, we have presented a model that is a step towards the establishment of patch diversity baselines for management purposes. The results derived from this model can be used: 1) to compare different landscapes in terms of diversity as an aid to selecting appropriate conservation and preservation areas; 2) to plan forest management strategies for biodiversity conservation; 3) to evaluate the effects of a change in the fire cycle; and, 4) to evaluate the effects of past and present forestry practices on levels of landscape diversity. Assuming that patch diversity reflects the other levels of biodiversity (genetic and species; Franklin, 1993), these types of data can be used to find out which landscape is most likely to show the highest natural level of biodiversity in order to select landscapes for conservation purposes. For instance, the results shown on Figures 4 and 5 are useful in comparing different landscapes in terms of natural patch diversity. The results can also be useful in defining forest management plans that maintain the patch diversity at a level similar to that expected in a natural landscape in quasi-equilibrium with a fire cycle. For example, the results in the typical landscape (L49) indicate that the natural distribution of patches would be composed of 31% of coniferous forest, 15% mixed-coniferous forest, 14% mixed-deciduous forest, and 38% deciduous forest. Thus, with these data, an expected natural diversity index of patches of 0.96 is established and can serve as a baseline for management purposes. A manager might plan logging strategies that would maintain similar proportions of each forest type. Furthermore, by using the expected age-class distribution of forest types (Figure 4), one could

430

SYLVIEGAUTHIERET AL.

define a logging schedule in the landscape that would maintain the proportion of age-class-forest type at similar levels to that observed under natural conditions. The effects of past and present forestry practices, such as fire suppression, on forest diversity in a landscape can also be evaluated. For instance, the expected proportions of forest types can be compared to actual proportions using either forest inventory maps or remote sensing data. The fire suppression effect on forest cover composition and patch diversity can also be evaluated by changing the fire cycle length in the model, as in Figure 6. The results indicate that, by suppressing fire, patch diversity would decline in all landscapes. For instance, our results showed that, in a typical landscape (L49), the patch diversity level with fire suppression is decreased to 81.9% of the level expected under a natural fire cycle. Moreover, the results suggest that all landscapes do not have the same sensitivity to fire suppression; one district showed very little decrease in patch diversity (district C22, Figure 6). Therefore, the modelling of forest type dynamics with varying fire cycles can also be useful in comparing landscapes in terms of sensitivity to climate and fire cycle changes. 5.2. FUTURE DEVELOPMENTS

Several improvements would make the model more powerful. The model presented here is based on empirical data. The use of this type of data reduced the prediction ability of the model and its future generalization potential. Factors such as prefire composition, fire size and intensity, and fire-to-fire interval were not taken into account. For instance, coniferous tree establishment is highly related to the abundance of seed sources. This abundance may vary as a function of fire size, time since fire, intensity or season (Fryer and Johnson, 1988; Rebertus et al., 1991). Pre-fire stand composition is also very important in explaining post-fire succession pathways. An example comes from jack pine forests, where time since fire can only predict the species abundance in stands where jack pine was present prior to the fire (Bergeron and Dansereau, 1993). In fact, if there is a long enough interval between two successive fires, jack pine is likely to disappear from the site because it has mainly closed cones and cannot easily reinvade from a distance into the burned area (Gauthier, 1991). The use of more mechanistic models such as those described by Shugart (1984), Prentice and Leemans (1990), and Fulton (1993) may improve the predictive capacity of this type of modelling. However, gap-models were not designed to simulate the landscape where large-scale disturbances are the driving force of forest dynamics. The development of more mechanistic models at the landscape level, in a spatially-explicit context, will permit the inclusion of chance events such as pre-fire vegetation composition and seed source distance in modelling post-fire succession. The model presented in this paper focused only on fire disturbance, despite the fact that insect outbreaks are also important in the boreal forest (Stocks, 1985; Bonan and Shugart, 1989). Insect outbreaks, like spruce budworm, would be like-

FOREST DYNAMICS, FIRE CYCLES, AND NATURAL MOSAIC DIVERSITY

431

ly to reintroduce deciduous species within the old coniferous forests. However, Bergeron and Dansereau (1993) suggested that with a fire cycle < 200 years forest composition of the region is mainly driven by post-fire succession while, for longer fire cycles, mixed-deciduous forest would be maintained by spruce budworm. The inclusion of other types of disturbances concurrently with the use of a more mechanistic model would make our model more powerful. The theoretical age-class distribution of each landscape is based on the assumption of quasi-equilibrium. At the landscape level, depending on the size of the area under study and the disturbance regime, vegetation may or may not be in quasiequilibrium with the disturbance cycle (Shugart, 1984) and some evidence suggests that fire-prone landscapes rarely reach a quasi-equilibrium state (Romme, 1982; Baker, 1989). Moreover, several studies in different areas, including the study area, have shown that the fire cycle has changed within the last 300 years (Clark, 1988; Bergeron, 1991; Engelmark et al., 1994). Thus the vegetation is likely to have encompassed several fire cycle changes. However, the use of a simplistic age-class distribution provides a general picture of the landscape fire regime and age-class distribution, allowing comparisons of landscapes in qualitative or semi-quantitative terms (Turner and Romme, 1994). The quasi-equilibrium state is also affected by the size of a landscape (Turner et al., 1993). For instance, in the ecological region studied, the average area of a landscape (district) is relatively low (37 000 ha; Table I) while the area burned in a single fire can exceed that size (Heinselman, 1981; Stocks and Simard, 1993). Thus, managing at the landscape (district) level based on the results of the model could increase patch diversity. Studies on the spatial and size distribution of natural fires must be undertaken to see if the vegetation is in quasi-equilibrium with the fire cycle (Johnson, 1992). On the other hand, the development of spatially-explicit models would avoid this problem, because these models do not require assumptions about the quasi-equilibrium of the landscape (Turner and Romme, 1994).

6. Conclusion

To assist managers in answering questions such as where and when to proceed with an intervention it will be essential to develop a spatially explicit model. The spatial distribution of patches is also important in a landscape for conserving biodiversity. For instance, the post-fire establishment of plants that depend on seed dispersal is affected by the size of the fire, the distance to seed sources in unburned sites, and the fire severity (Turner and Romme, 1994). Furthermore, it is recognized that birds and mammals need different habitat types side-by-side to maintain their population levels (Noss, 1983). The framework of natural disturbances provides an ecological basis to manage our forest landscapes sustainably. The model presented here, despite its simplicity, has the potential to help a forest manager in planning intervention in landscapes. The development of a spatial landscape model, to be

432

SYLVIEGAUTHIERET AL.

jointly used with m o r e mechanistic models, will give better predictive ability to this type o f modelling. Such models, used in conjunction with ecological m a p s and ecological classifications (such as the one o f the Minist~re des Ressources naturelles du Qu6bec), will result in more powerful tools for sustainable forest management.

Acknowledgements We thankfully a c k n o w l e d g e the financial support o f the Federal G o v e r n m e n t Green Plan (Forestry Practices, Fire and Decision Support S y s t e m Initiatives; SG) and o f N S E R C funding (YB). The support of the Minist~re des Ressources naturelles du QuEbec is also acknowledged. We thank Danielle Charron and France Conciatori for field assistance. We also wish to thank Steen Magnussen, Ian T h o m p s o n , M i k e Flannigan, R o b McAlpine, A s o k a Yapa, Mike Wotton, and two a n o n y m o u s reviewers for n u m e r o u s c o m m e n t s and advice on this manuscript.

References Attiwill, EM.: 1994a, 'The disturbance of forest ecosystems: the ecological basis for conservative management', For. Ecol. Manage. 63, 247-300. Attiwill, P.M.: 1994b, 'Ecological disturbance and the conservative management of eucalypt forests in Australia', For. Ecol. Manage. 63, 301-346. Baker, W.L.: 1989, 'Landscape ecology and nature reserve design in the Boundary Waters Canoe Area, Minnesota', Ecology 70, 23-35. BEland, M.: 1990, 'Le cadre 6cologique forestier du QuEbec: rEalisation et applications dans un secteur de l'Abitibi', M.Sc. Dissertation, Universit6 du QuEbec ~ MontrEal, QuEbec. Bergeron, Y.: 1991, 'The influence of island and mainland lakeshore landscapes on boreal forest fire regimes', Ecology 72, 1980--1992. Bergeron, Y., Bouchard, A., Gangloff, P. and CamirE, C.: 1983, 'La classification Ecologique des milieux forestiers de la partie ouest des cantons d'HEbEcourt et de Roquemaure, Abitibi, QuEbec', Etudes 6cologiques 9, Laboratoire d'Ecologie foresti~re, UniversitE Laval, QuEbec. Bergeron, Y. and Bouchard, A.: 1984, 'Use of ecological groups in analysis and classification of plant communities in a section of western QuEbec', Vegetatio 56, 45-63. Bergeron, Y. and Dansereau, R.-P.: 1993, 'Predicting the composition of Canadian southern boreal forest in different fire cycles', J. Veg. Sci. 4, 827-832. Bonan, G.B. and Shugart, H.H.: 1989, 'Environmental factors and ecological processes in boreal forests', Annu. Rev. Ecol. Syst. 20, 1-28. Canadian Forest Service: 1993, The State of Canada's Forests 1993, Natural Resources Canada, Ottawa, Ontario. Carleton, T.J. and Maycock, P.F.: 1978, 'Dynamics of the boreal forest south of James Bay', Can. J. Bot. 56, 1157-1173. Clark, J.S.: 1988, 'Effects of climate change on fire regime in northwestern Minnesota', Nature 334, 233-235. Cogbill, C.V.: 1985, 'Dynamics of the boreal forests of the Laurentian Highlands, Canada', Can. J. For. Res. 15, 252-261. Dansereau, R.-P. and Bergeron, Y.: 1993, 'Fire history in the southern boreal forest of northwestern Quebec', Can. s For. Res. 23, 25-32. Engelmark, O., Kullman, L. and Bergeron, Y.: 1994, 'Fire and age structure of scots pine and norway spruce in northern sweden during the past 700 years'. New Phytol. 126, 163-168.

FOREST DYNAMICS,FIRE CYCLES, AND NATURALMOSAICDIVERSITY

433

Forman, R.T.T. and Godron, M.: 1986, Landscape Ecology, Wiley & Sons, New York, USA. Franklin, J.F.: 1993, 'Preserving biodiversity: species, ecosystems, or landscapes', Ecol. Appl. 3, 202-205. Fryer, G. I. and Johnson, E.A.: 1988, 'Reconstructing fire behavior and effects in a subalpine forest', J. Appl. Ecol. 25, 1063-1072. Fulton, M.R.: 1993, 'Rapid simulation of vegetation stand dynamics with mixed life-forms', In: A.M. Solomon and H.H. Shugart (eds.), Vegetation Dynamics and Global Change, Chapman & Hall, New York, pp. 251-271. Gauthier, S.: 1991, 'Structure g6n6tique et s~rotinisme de populations de pin gds (Pinus banksiana Lamb.) soumises ~ deux r6gimes des feux distincts', Ph.D. Thesis, Universit6 de Montr6al, Qudbec. Hansen, A.J., Spies, T.A., Swanson, F.J. and Ohmann, J.L.: 1991, 'Conserving biodiversity in managed forests: lessons from natural forests', BioScience 41, 382-392. Heinselman, M.L.: 1981, 'Fire and succession in the conifer forests of northern North America', In: D.H. West, H.H. Shugart and D.B. Botkin (eds.), Forest Succession: Concept and Applications, Springer-Verlag, New York, pp. 375--405. Johnson, E.A.: 1992, Fire and VegetationDynamics: Studiesfrom the NorthAmerican Boreal Forests, Cambridge Studies in Ecology, Cambridge University Press, New York, 129 pp. Mladenoff, D.J., White, M.A., Pastor, J. and Crow, T.R.: 1993, 'Comparing spatial pattern in unaltered old-growth and disturbed forest landscapes', Ecol. Appl. 3, 294-306. Noss, R.E: 1983, 'A regional landscape approach to maintain diversity', BioScience 33, 700--706. Noss, R.F.: 1987, 'Protecting natural areas in fragmented landscapes', NaturalAreas Journal 7, 2-13. Pickett, S.T.A. and White, P.S.: 1985, The Ecology of Natural Disturbance and Patch Dynamics, Academic Press, New York. Pimentel, D., Stachow, U., Takacs, D.A., Brubaker, H.W., Dumas, A.R., Meaney, J.J., O Neil, J.A.S., Onsi, D.E. and Corzilius D.B.: 1991, 'Conserving biological diversity in agricultural/forestry systems', BioScience 42, 354-362. Prentice, I.C. and Leemans, R.: 1990, 'Pattern and process and the dynamics of forest structure: a simulation approach', J. Ecol. 78, 340-355. Rebertus, A.J., Burns, B.R. and Veblen, T.T.: 1991, 'Stand dynamics of Pinus flexilis-dominated subalpine forests in the Colorado Front Range', J. Veg. Sci. 2, 445--458. Robitaille, A.: 1989, 'Cartographie des districts 6cologiques: normes et techniques', Minist~re de rEnergie et des Ressources, Service de l'inventaire forestier, Division 6cologique, Qu6bec. Romme, W.H.: 1982, 'Fire and landscape diversity in subalpine forests of Yellowstone National Park', Ecol. Monogr. 52, 199-221. Rowe, J.S.: 1972, Forest Regions of Canada, Environ. Can., Ottawa, Ontario. Shannon, C.E,: 1948, 'A mathematical theory of communication', Bell System Tech. J. 27, 379-423. Shugart, H.H.: 1984, A Theory of Forest Dynamics, The Ecological Implications of Forest Succession Models, Springer-Verlag, New York. Sims, R.A., Towill, W.D., Baldwin, K.A. and Wickware, G.M.: 1989, Field Guide to Forest Ecosystems Classification for Northwestern Ontario, Ont. Min. Nat. Res., Toronto, Ontario. Stocks, B.J.: 1985, 'Forest fire behavior in spruce budworm-killed balsam fir', In: Recent Advances in Spruce Budworm Research Proc. CANUSA Spruce Budworm Res. Symp., Sept. 16-20, 1984, Bangor, ME., Govt. Can., Can. For. Serv., Ottawa, Ontario, pp. 188-199. Stocks, B.J. and Simard, A.J.: 1993, 'Forest fire management in Canada', J. Disaster Manage. 5, 21-27. Turner, M.G. and Romme, W.H.: 1994, 'Landscape dynamics in crown fire ecosystems', Landscape Ecol. 9, 59-77. Turner, M.G., Romme, W.H., Gardner, R.H., O Neill, R. and Kratz, T.K.: 1993, 'A revised concept of landscape equilibrium: disturbance and stability on scaled landscapes', Landscape Ecol. 8, 213-227. Van Wagner, C.E.: 1978, 'Age-class distribution and the forest fire cycle', Can. J. For. Res. 8, 220-227.

434

SYLVIEGAUTHIERET AL.

Viereck, L.A.: 1983, 'The effects of fire in black spruce ecosystems of Alaska and northern Canada', In: R.W. Wein and D.A. MacLean (eds.), The Role of Fire in Northern Circumpolar Ecosystems, John Wiley & Sons, New York, pp. 201-220. Vincent, J.S. and Hardy, L.: 1977, 'L'6volution et I'extinction des lacs glaciaire Barlow and Ojibway en territoire qu6b6cois', Gdogr. Phys. Quat. 31, 357-372.