Simulation Modeling of the Effects of Site Conditions and Disturbance History on a Boreal

Forest Landscape
Author(s): Mark R. Fulton
Source: Journal of Vegetation Science, Vol. 2, No. 5 (Dec., 1991), pp. 603-612
Published by: Blackwell Publishing
Stable URL: http://www.jstor.org/stable/3236171
Accessed: 14/05/2009 11:55

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=black.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

Blackwell Publishing is collaborating with JSTOR to digitize, preserve and extend access to Journal of
Vegetation Science.

http://www.jstor.org
Journal of Vegetation Science 2: 603-612, 1991
? IAVS;Opulus Press Uppsala. Printed in the UnitedStates of America 603

Simulation modeling of the effects of site conditions and disturbance

history on a boreal forest landscape

Fulton, Mark R.

Department of Ecological Botany, Uppsala University, Box 559, S-75122, Uppsala, Sweden;
Tel. +46 18 2850; Fax +46 18 553419

Abstract. Computersimulationswere used to elaboratehypo- Introduction

theses aboutcontrols on forest structureand composition in a
0.7 km < area of boreal forest in CentralSweden. DBH and The 'special theory' of forest dynamics as articu-
species of all adult trees and stand conditions were recorded lated by Shugart (1984) has been widely applied to
for 57 - 10 m radiusplots. Ordinationof these data suggested
explain trends and patterns of forest structure and com-
thatnutrient-availabilityand time-since-disturbancewere the
main controls of forest composition and structurewithin the position. This theory, based ultimately on the ideas of
area. Watt (1947), considers forest dynamics over a time
The simulationmodel couples equationsrepresentingthe scale of decades to centuries to be a consequence of
effect of tree canopy structureand biomass on light and soil interactions between local populations of adult trees
conditions with equations representingthe effect of these and the environment of a small (ca. 0.1 ha) patch of
conditions on reproduction,growth and mortality in height ground. Thus, the essential mechanisms included are
cohortsof trees on a 0.1 ha patch.Nitrogen-availabilitylevels those of neighborhood competition (Mack & Harper
for each modeled plot were simulated by species-specific 1977). Differences between forest stands can be ac-
growthmultipliers.The model was runfor 400 simulatedyr at counted for by variations in boundary conditions over
six levels of N availability.Age and N status of each study
time or space (e.g. disturbance history or soil condi-
plot were inferredby matchingwith the most similar model
output.Inferredages agreed with what is known of the dis- tions), or as a result of partially stochastic processes
turbancehistory, and site factors relatedto soil fertility were such as mortality and establishment. The theory has
correctlycorrelatedwith the inferredN status. been embodied in simulation models of the 'gap model'
The consequences of size-selective disturbancewere ex- type (Botkin, Janak & Wallis 1972; Shugart 1984) and
plored by model experiments. Biomass was removed from successfully tested on a wide variety of forest types,
large or small size classes at 100 - 200 yr and the simulations including boreal forests (Bonan 1989; Prentice &
were runfor an additional300 yr. Disturbedstandsof high N Leemans 1990).
statusoften became similarto undisturbedstandsof different
Because gap models incorporate autogenic proc-
N status. Size-selective disturbancesproduced stands that
were different from any in the undisturbedsuccession, but esses, such as local competition for light, the response
these differences disappearedwithin 50 - 100 yr, implying of species to a gradient of site conditions can be time-
successionalconvergence in standstructureandcomposition. dependent (Smith & Huston 1989). The physiological
Plots of simulatedbasal areaagainsttime andnitrogen-avail- responses of the species to the gradient can be com-
ability for the four species illustratethe time dependenceof pared with the predicted species performance at any
species performancealong a fertilitygradient. given time. This allows these models to predict differ-
ences between the fundamental niche and the realized
niche - a fundamental issue in gradient analysis.
Keywords: Canopy structure;CentralSweden; Cohort; For- The purpose of this paper is to apply the 'special
est dynamics;Gap model; Nitrogen availability;Ordination. theory' to account for differences in species composi-
tion and size structure between sites in a boreal forest
landscape in Central Sweden. Simulation exercises were
Nomenclature: Lid (1985).
carried out to elaborate hypotheses about the causes of
forest compositional and structural variations observed
within the study area today. The model was further
used to clarify the possible role of size-selective distur-
bances on the accuracy of ecological calibration, and to
604 Fulton, M. R.

illustrate the behavior of the major species over time in Table 1. Basic equationused in FLAM.
relation to a fertility gradient.
(1) Light intensity at depthz in canopy.

Iz = Io exp (-kFz)
The study area and preliminary data analysis

Io = light intensityat top of canopy; k = extinctioncoeffi-

The study area is located in a mixed-conifer forest in cient; Fz = cumulativeleaf area index at depthz.
Tiveden National Park in South Central Sweden. The
landscape consists of glacial till of variable thickness (2) Net photosynthesis/leafarea at depthz:
over granitic bedrock. Spruce (Picea abies) dominates
in ravines with thick mineral soil and pine (Pinus
klz -c
sylvestris) dominates on bogs and rocky upland areas. Z zp kI-
kI +o-c c
Birch (mostly Betula pubescens with some B. pendula)
and aspen (Populus tremula) are found in smaller
c = compensationpoint; a = half-saturationpoint.
amounts throughout. Much of the area has been periodi-
cally cut for charcoal from the 16th century through the (3) Soil competitionmultiplier:
19th century (Back & Renstr6m 1988).
In 1982, the National Swedish Environmental Pro- C =(Wmax - W) / Wmax
tection Board (SNV) established a monitoring site in the
western area of the park. Data on species composition W = patch biomass; Wmax= maximum patch biomass.
and dbh of adult (>2 cm dbh) trees were collected by
SNV workers for 57 10-m radius circular plots located (4) N-availabilitymultiplier:
on a grid in a 0.7 km2 rectangular area. A description of
the area and analysis of these data in terms of forest and
N=r d+ae(l10-c
ground vegetation types may be found in Back & (b+4AN-170))
Renstr6m (1988). The data were compiled into a matrix
of basal area (m2/ha) of each species in 5 diameter
classes (2.0-5.9 cm, 6.0-13.9 cm, 14.0-23.9 cm, 24.0- AN = Available Nitrogen (kg ha-l yr-l); a, b, c, d, e =
36.0 cm, >36.0 cm). Environmental data included soil parametersset accordingto N-defiency tolerance;r = factorto
texture class, soil type, mean depth of mineral soil, rescale N to 1 at AN = 65.
slope, aspect, and site moisture class. Age distributions
of trees could not be collected because tree coring is (5) Growthrate:
prohibited in monitoring sites, but the historical infor-
mation summarized in Back & Renstr6m (1988) sug- A(D2H) H
NCB Sl (yPz 6)d
gests that most stands are 100-200 yr old. year
A two-dimensional ordination of the 57 plots was
carried out by local non-metric multidimensional scal-
D2H = diametersquaredx height;Sl = verticaldensityof leaf
ing on the basal area/size class data (Prentice 1980; area;y= growthscalingconstantfor species;83= maintenance
Minchin 1987) using the Bray-Curtis dissimilarity cost of leaves at height z; B = bole height of tree (leafless
measure: trunk).

2min(x ij,xk)
dIk = 1 - (1) (6) Height as a functionof diameter:
Xijl +Xijk

where dlkis the dissimilarity between plots I and k, and H = 1.3 + (Hmax- 1.3) (1 - exp(-aD))

X,jk is the basal area of species i in size class j of plot k.

Each size class of each species was treated as a separate Hmax= maximumheight for species; a = constantfor species.
'species' so that the resulting ordination reflected both (7) Relative growthefficiency:
structural and compositional differences between plots.
The Bray-Curtis measure has been recommended as a
E r = A(D 2H)
good general measure of ecological distance (Beals yLPo
1984; Faith, Minchin & Belbin 1987). A DECORANA
ordination (Hill & Gauch 1980) was used for the initial L = leaf area; Po = Pz evaluated at Iz = Io
configuration, as recommended by Minchin (1987).
Patterns related to the recorded environmental variables
 - Simulationmodeling of a boreal forest landscape- 605

Spruce BA Pine BA BA/Tree

0 0 *
0
4
* * 0
4a 0 0 % 0
00 . 0
. . .1 0 00 .p0 *
0 , *
0
0

0
a**e.. *

*
0.

..6 0
* *

* - 0.0 m2 * - 0.0 m2 . - 0.01 m2/tree

* - 1.3 m2 *- 1.2 m2 * - 0.10 m2/tree

.
Podsol Soil Peat Soil Bare Rock

. 0
0
0
* 0 @* . 0
* . .0 .
0 * b 00 0 .0 * *
9 .4w. *..'
* 0.
0 0
*
* .

Fig. . A non-metric
multidimensional of the57 fieldplots;showingoverlaysof pinebasalarea,sprucebasalarea,
scalingordination
mean basal area/tree,and dominantsoil type.

were found by graphicaloverlays, some of which are may preferentiallyfell the larger trees in a stand. The
shown in Fig. 1. accuracy of ecological calibration,where site condi-
The analysisshowed severalpatterns:(1) the sites on tions are inferred from the vegetation of the site
both bogs and rocky upland sites had sparse, pine- (Jongman,ter Braak& van Tongeren1987; ter Braak&
dominatedforests, (2) the environmentalfactors most Prentice 1988), can be affected by these kinds of dis-
clearlyrelatedto the species compositionwere soil type turbance.Simulationexercises were used to assess the
(podsol vs. peat soil) and depthof mineralsoil, and (3) possible effects of size-selective disturbanceson the
the remainingvariationwas relatedto the relativedistri- apparentage and N-status of stands, and to suggest
butionof basal area in the differentsize classes. These some possible stabilitycharacteristicsof forest succes-
patternssuggested that the main controls on the forest sion in this system.
may be relatedto nutrientavailabilityand stand age. A
model exercise, including plots of different nitrogen
availability, is used here to test the sufficiency (cf. Methods
Botkin 1981) of this hypothesis.Rocky uplandsites and
sites with peaty soils should resemble simulationswith Simulation model
low nitrogenavailability,andsites with thickerpodsolic
soils should resemble simulationswith higher nitrogen The model FLAM (ForestLAyer Model) simulates
availability. the numbers, leaf area and stem volume of trees in a
Some partsof the forest may have been affected by small numberof height classes on a small (ca 0.1 ha)
size-selective removal of adult trees. Cuttingfor char- patch (Fulton in press a). For these model runs, the
coal burningsometimesinvolved only removal of trees patch was representedby ten 4 m-thick height cohorts
in the smallersize classes (S. Brakenhielmpers.comm.); from 1.3 m to 41.3 m. Equationsrepresentingthe effect
the largetrees were left eitheras seed sourcesor because of the trees on the local light and soil environmentare
they areunsuitablefor charcoalproduction.Windstorms linkedwith equationsgoverningthe recruitment,growth
606 Fulton, M. R.

area,and is decreasedby a fixed annualproportion(the

sapwood turnover rate) and by pruning of leaf area
shaded below the species compensationpoint. Fulton
(in press a) describes the applicationof these relation-
ships to a height-class structuredmodel. The mortality
rate is a step function of relative growth efficiency
(Prentice& Leemans 1990); if the efficiency falls be-
low a thresholdthe probabilityof mortalityis consider-
ably enhanced.
Trees are introduced at a height of 1.3 m by a
stochastic function of the environment.The expected
numberof saplings is proportionalto the growthmulti-
plier for the species calculatedfor conditionsunderthe
Fig. 2. The randomizationtest of model performanceshown canopy. The assumption implicit in this relationship,
schematicallyas a Venn diagram.The test comparesthe data/ that the density of recruitmentof small trees is posi-
model overlap with the overlap between the model and ran-
tively correlatedwith their growth rate, was borne out
domly selected points within the test space.
by observations in the study area, although the exact
relationshipis not linear (Fulton in press b). A similar
formulationfor tree recruitmentis used by Pastor &
and mortality of adult trees on the patch (Table 1). Post (1985).
Recruitmentand mortalityare stochastic,so the behav- To allow direct comparisonwith the field data, the
ior of a stand is simulated by the average of several model provideda descriptionof each patch in termsof
replicatepatches. basal area (m2/ha) in five diameter classes for each
Adult trees affect the environment of the model species.
patch and this environmentin turn affects the trees by
multipliersthat decrease the growth rate from the spe- Model exercises
cies-specific maximum.The Beer-Lambertlaw (Monsi,
Uchisima & Oikawa 1973) relatesthe decrease of light Threemodel exercises were carriedout. Parameters
through the canopy to the vertical distributionof leaf for pine, spruce, birch (B. pubescens) and aspen were
area.The shadingof leaves modifies growththroughan the same as in Prentice& Leemans (1990), except that
asymptotic equation whose parametersare the photo- the maximumrecruitmentratesof birchand aspenwere
synthetic compensationpoint and half-saturationpoint lowered to 5/ha/yr and 0.5/ha/yr respectively because
for the species. Prentice & Leemans (1990) discuss these species areheavily affectedby browsingby moose
how these parametersmay be derived or approximated and roe deer in Tiveden (Back & Renstr6m 1988 and
for this modeling context. Symmetric competition pers. obs.).
(Weiner & Thomas 1986) for soil resources is simu- The first exercise was run for comparisonwith the
lated by specifying a maximumbiomass for the patch, field data. The protocol for data-model comparison
and tree growth is decreased as this maximum is ap- closely resemblesthatdescribedby Harrison& Shugart
proached. (1990). First, the model was run to produce a set of
A fixed nitrogen-availability(kg/ha/yr)was speci- stands varying in age and N-availability. The age and
fied for each modeled stand.Each species was assigned N-status of each of the 57 field plots was inferredby
a nitrogentoleranceclass (Helmisaari& Nikolov 1989), matchingto the most similar model stands;a form of
and equations relating nitrogen availability to growth ecological calibration. The resulting distributionsof
response (Aber & Melillo 1982) were appliedto derive inferredN-status and age were then comparedto the
growthmultipliers.These factorswerere-scaledto equal availableinformationabout the field plots as a test for
1.0 at a nitrogen-availabilityof 65 kg/ha/yr. The con- the adequacyof the model.
jecture underlyingthis static N-availability is that the The specifics of the comparisonwere as follows. 20
between-site variationsin this parameterare more im- replicatepatcheswere runfor 400 yr at each of six levels
portantthan the within-sitevariationsover time. of N-availability, from 10 to 78 kg/ha/yr. Ellenberg
Simulated tree growth, as an increment of stem (1971) reportsN-mineralizationratesof 13-79 kg /ha/yr
volume, is distributedaccordingto a function relating for cold-temperatespruce forests. The diameter-class
height to diameter. In keeping with the pipe-model distributions(averagesforthe 20 replicatepatches)were
theory of tree function (Waring, Schroeder & Oren outputevery 20 yr. TheBray-Curtissimilarityindexwas
1982), leaf areaincreaseslinearlywith increasingbasal used for matching field plots with model stands.Each
 - Simulation modeling of a boreal forest landscape - 607

size class of each species was treated as a separate gradient.100 replicatesat each of 16 levels of N-avail-
'species' so thatboth structureand species composition ability,from 5 to 80 kg/ha/yr,were runfor 400 yr. Total
were matched.The inferredage andinferredN-statusof basal area(m2/ha)was recordedfor each species at 20 yr
each field plot were then comparedwith the environ- intervals.To furtherillustratehow the model distributes
mental field data using Kendall's r correlationcoeffi- the species along the fertility gradient,the correlation
cient. The distributionof inferredages could only be between the values of the N-availability growth re-
informally comparedto the disturbancehistory of the sponse multipliers and the basal area of each species
areadue to the lack of precise standhistories. was calculatedfor each recordedyear.
Because of the indirect nature of the data-model
comparisons,an additionalmeasureof model perform-
ance was applied.The test describedhere was designed Results
to determinewhetherthe model outputis more similar
to the field data than to a randomly generated set of Comparisonwithfield data
configurations.Given that the model output and the
field dataeach comprisea subsetof all 'possible' stands, The inferred ages of the field plots were mostly
one reasonable measure of model performanceis to from 100 to 200 yr. InferredN-status was generally
compare the data-model similarity with the similarity from 24 to 65 kg/ha/yr (Table 2). The matched model
between the model and a randomsample of 'possible' stands from 200-380 yr were of mostly of low N-
stands. The outcome of such a test is affected by the status. As expected, soil texture class and podsol soil
rangeof standsconsideredto be possible; the largerthe type were positively correlated with the inferred N-
range, the more significantthe data-modelsimilarityis status (Table 3). Peat soil and bare rock were nega-
madeto appear(unlessthe model has no similarityto the tively correlatedwith inferredN-status, but the latter
field data at all). A fairly rigorouscriterionwas chosen correlation was not significant. No site factors were
for this study, using the limits of the field data alone. A significantly correlated with inferred age. The mean
test space was delimited between the maximum and similarityof the field plots to the nearest model stand
minimumbasal areasfor each species in each diameter was 62.7%, comparedto 56.3% for the 1000 randomly
class from the field data (Fig. 2). With 4 species and 5 generated plots (difference significant at p < 0.001).
diameterclasses, the resultingspace has 20 dimensions. The similarityof field plots to the nearestmodel stand
1000 samples were drawn at randomfrom this space was correlated with the inferred age (Kendall's T =
andthe Bray-Curtissimilarityto the nearestmodel stand 0.288, p < 0.005).
was noted for each. The resultingdistributionof simi-
laritieswas comparedwith the distributionof field data-
model similarities. Table 2. Number of field plots matchedto models stands of
The second exercise simulatedthe effects of partial differentage and N-availability.
disturbanceson the developmentof stands. The patch
descriptionsfrom the first model exercise were retained N-availability
from 100 and 200 yr. Each set of patches(at each of the (kg/ha/year)
Row
six levels of N-availability)was subjectedto two types 10 24 37 51 65 78 totals:
of disturbance:(1) 'low-grading'where 80%of the total
biomasswas removedstartingwiththe smallsize classes, 80 2 2
and (2) 'high grading' where 80% of the total biomass 100 2 2 2 6
120 6 2 1 9
was removed startingwith the large size classes. The 140 3 3 3 9
resultingfour treatmentswere runfor an additional300 160 1 2 1 2 1 7
yr, andthe diameterclass distributionsfrom every 20 yr 180 1 1 1 3
were matchedwith standsfrom the first model exercise 200 1 3 1 2 7
Model 220 2 1 3
as in the comparisonof field andmodel dataabove. One Year 240 2 2
additionalN-availability,92 kg/ha/yr,was addedto the 260 1 1
first model exercise to extend the range of possible 280 3 3
calibrations.The Bray-Curtissimilarityof the disturbed 300 1 1
320 2 2
stands to the nearest stand from the first exercise was 340 0
noted, as well as the inferredage and N-status of the 360 1 1
disturbedstands. 380 1 1
The third exercise was performedto examine the
Columntotals: 1 C11 11 16 11 7
time-dependenceof species' responses to the fertility
608 Fulton, M. R.

Year 160 Site Al

Pine Spruce Birch Aspen Pine Spruce Birch Aspen
>36 cm >36 cm
24-36 cm 24-36 cm
14-24 cm 14-24 cm I
6-14 cm 6-14 cm
2-6 cm 2-6 cm

0 20 0 20 0 10 0 10 0 20 0 20 0 10 0 10
m2/ha m2/ha m2/ha m2/ha m2/ha m2/ha m2/ha m2/ha

Year 160 Site H3

Pine Spruce Birch Aspen Pine Spruce Birch Aspen
>36 cm I >36 cm
24-36 cm 24-36 cm
14-24 cm
6-14 cm
2-6 cm _ _
14-24
6-14
2-6
cm
cm
cm
' p
r
0 20 0 20 0 10 0 10 0 20 0 20 0 10 0 10
m2/ ha m2 /ha m2 / ha m2/ha m2/ha m2/ha m2/ha m2/ha

Year 160 Site F4

Pine Spruce Birch Aspen Pine Spruce Birch Aspen
>36 cm I >36 cm
24-36 cm 24-36 cm F
14-24 cm
6-14 cm
2-6 cm
14-24
6-14
2-6
cm
cm
cm
I
0 20 0 20 0 10 0 10 0 20 0 20 0 10 0 10
m2/ ha m2 / ha m2/ha m2/ha m2/ha m2/ha m2/ha m2/ha

Fig. 3. Structuralprofilesfor simulationoutputat 160 yr, andmatchingfield plots. Profilesshow the basalareaha-' for each size class
and species. The top, middle and bottom simulatedstandshave N-availabilitiesof 24, 51 and 78 kg ha-1yr-1respectively.

Diagramsof model standsfrom simulatedyear 160 than the tolerancelimit of spruce.This was apparently
matched with field plots (Fig. 3) illustrate predicted due to a close matchbetween the size structuresof pine
stand structures at different N-availabilities and the in the model and field plot.
degree of residualvariationnot capturedby the model.
Generalqualitativeand quantitativeagreementis high, Partial disturbance
but the model tends to overestimate the presence of
birchand aspen. In a few cases, a field plot with spruce The response of stands to partial disturbancede-
present was matched to a model plot with lower N-status pended in a complex fashion on the N-status and the
specific type of disturbance.The time for disturbed
stands to become indistinguishablefrom undisturbed
standswas 100-150 yr for the low gradingdisturbances
Table 3. Non-parametriccorrelations(Kendall'sr) of inferred and 50-100 yr for the high grading disturbances(Fig.
standage andN-availabilitywith site variablesrecordedin the 4). The low gradingdisturbancesproducedmore atypi-
field. cal structuresand compositions, and the stands took
longer to re-convergeto the 'normal'course of succes-
Model Year Model N-availab. sion. High graded stands of low N-status (< 51 kg/ha/
yr) were generally indistinguishablefrom undisturbed
Soil textureclass -0.134 (n.s.*) 0.439 (p < 0.001) stands.
Podsol soil type -0.152 (n.s.) 0.406 (p < 0.001)
Peat soil type 0.127 (n.s.) -0.359 (p < 0.005)
Disturbedstandswith N-statushigherthan24 kg/ha/
Abundantbarerock -0.089 (n.s.) -0.150 (n.s.) yr were frequently matched with undisturbed stands of
different N-status. The low grading at 100 yr caused the
n.s.* = implies p> 0.1.
more fertile disturbed stands to have a lower inferred
 - Simulationmodeling of a boreal forest landscape 609

78- -+++********** 78- 0+ - - - - - - 0--00- - 78-1111 .... ..111 ?

Lowgrading N-availability of 65---++++ * * * * 65 - ... - - - - -- - 65 -El ?sass.._ *mE**
at 100 years disturbed patches 51 - - - + + + + * * * * 51- .-- - - 00 000000 51- *-*****U.**.***
(kg/ha/year) 37 - - - - * * * * * * * 37 0000000000000000 37 m uumm
m E m ?
24- - ++ * * * * 24 - - - + 0 0 0 0 O0 0 0 0 0 0 0
10 + + + + * 10 -0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1024-m---- .........
6 100 200 300 0 100 200 300 6 100 200 300

78- ++++********* 78 0 0 0 0 0 + 0 + . . . . . . . . 78 - _ . a** a**

Lowgrading N-availability of 65 - - - - * * * * * * * * * 65 + 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 65 _u ? w*"mu*-*mm.mE
at 200 years disturbed patches 51- - * * * * * * * * * * 51 - - - - - - - - - - - - - - - 00
(kg/ha/year) 37 - - - - + * * * * * * * * * 37 - - - - - - 000 000000 37?--- ?*?Em
24- * * * * * * * * 24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00 24- *mi _m
10- --+++*********** 10 0000000000000000 10-. ? ?
mU.. * ***

0 100 200 300 0 100 200 300 6 100 200 300

78- +++************* 78 -0 - - 0 + O 0 0 0 0 0 0 0 0 0 0
High grading N-availability of 65 -+ ++ +* * * * * * * * * * * * 65 * 000+++++++++++ 65- _ -...-..
65 ** **r
at 100 years disturbed patches 51 - + + + + + * * * * * * + * * * * 51 *++++++++++++ 0 0 0 51 .?m u.
(kg/ha/year) 37 37 0-- v O 0 0 0 0 0 0 0 0 0 0
24 * 24 - 0 0 0 0 0 0 0 0 0 0 0 0 0 0 24- :- l U_-__E*? :
10 ++************** 10 0 0 0 0 0000000000 10 -- - - - -? - - - EE
0 100 200 300 0 100 200 300 0 100 200 300

78- * * * * * * * * * * * 78 - 0 - - - - 0 0 0 0 0 + + 0 + + 78 - .. ?*...?
.*?*?*
High grading N-availability of 65 * * * * * * * * * * 65 - * 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 65 - ?_uw*MMMu*MM*
at 200 years disturbed patches 51 + +* * * * * * * * * + 51- *+++++++++++++++ 51 - - - - - - - _ . n n n a 0* *
(kg/ha/year) 37- +* * 37- + 00 0 0 0 0 0 0 0 37 - ._,E*EM M
.*m----
24 *********** * 24- 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00 24 -. ... * g* **
10-**************** 10 - 0000000000000000 10 .- .... .*******

6 100 260 300 6 100 200 300 0 100 200 300

Years since disturbance Years since disturbance Years since disturbance

KEY KEY KEY

Bray-Curtis similarity of disturbed stands Disturbed stand matched with Age of most similar undisturbed stand
to most similar undisturbed stands: undisturbed stand of N-availability:
- 100 years
- < 0.5 * Two classes higher than disturbed stand * 200 years
+ 0.5 to 0.8 + One class higher than disturbed stand * 300 years
* 0.8 to 1.0 0 Same class as disturbed stand * 400 years
- One class lower than disturbed stand
= Two classes lover than disturbed stand

Fig. 4. Comparisonsbetween stand simulations with size-selective disturbancesand the original undisturbedmodel succession,
obtainedby calibratingthe disturbedstandsagainstthe originalmodel succession as for the field data.Each row of boxes gives the
resultsfrom a single disturbanceexperiment,and each column of boxes presentsa type of comparison.

N-status for as long as 300 yr, due to the selective Species responses to time and N-availability
removal of slower growing spruce in the understory.
The same effect occurredin stands of intermediateN- Each species had differentquantitativeand qualita-
status that were low graded at 200 yr. High grading tive responses to the time and N-status gradients(Fig.
selectively removedfastergrowingpines fromthe more 5). Spruce dominatedin the fertile end of the gradient
fertile stands, making the inferred N-status higher in and uniformly increased with time. Pine had attained
some cases. The standswith N-statusof 24-10 kg/ha/yr peak biomass in the middle of the gradientby the end of
were little affected in this way. In general the more the simulation; it was decreasing in the fertile end.
fertile stands have more contrastingoverstory and un- Aspen was most abundantbetween 100 and200 yr in the
derstoryspecies composition,so size-selective thinning most fertile part of the N-gradient. Birch was most
has a more noticeableeffect on the inferredN-status. abundantat about 100 yr and persisted only in the
The inferredage of the disturbedstandswas usually intermediatepart of the gradient.The basal area of all
lowered by disturbance,particularlyhigh grading.For fourspecies was well correlatedwiththe nitrogengrowth
some of the more fertile stands, the inferred age was responsecurves for the first 100 yr (Fig. 6); the correla-
raisedby low gradingbecause this disturbancemimics tion of the shade intolerantspecies began to decrease
the thinningof the understorycharacteristicof a mature rapidly after this time. This decreased correlationwas
stand. After re-convergenceto a 'normal' stand struc- most pronouncedfor the N-deficiency tolerantspecies
ture,the inferredage generallyincreasedin even steps. pine andbirch;the latterspecies had a negative correla-
tion by 260 yr.
610 Fulton, M. R.

Spruce
40

30 X S-- ---4-
Spruce Aspeli
Basal Area 20 \. -XL _
(m2/ha) Pine
00
\

-- Bilch
Years -0.5-
trogen Availability I I , . I . I I I
I ()( 200 300( 4(00
(kg/ha/year)
Yea-
40

Fig. 6. Correlationsbetween simulatedtotal basal areaandN

Pine
growthresponse curves for each species at 20 yr intervals.
Basal Area
(m2/ha)

Discussion

40
Nitrogen Availability In general, the model appears to account well for
(kg/ha/year) the range of forest structures and compositions in the
Tiveden study area. Inferred age and N-status are clearly
Birch distinguished due to the use of size classes in the data
Basal Area 5 model comparison, and the site factors are correctly
(m2/ha) correlated with the inferred N-status. The modal in-
ferred N-status falls in the middle of the range reported
by Ellenberg (1971). The distribution of inferred ages
80
is essentially as expected, except that there may be too
?'/60
40 many stands with a high inferred age; the historical data
20 Nitrogen Availability
(kg/ha/year) (Back & Renstr6m 1988) suggest that most of the area
10 has been cut or otherwise disturbed more recently than
200 yr. However, the results of the size-selective dis-
Aspen turbance exercises imply that the exact age and N-
Basal Area 5
status calibration of any particular stand should be in-
(M2/ha)
terpreted with caution.
The model leaves out several potentially important
80
mechanisms that may account for some of the varia-
60 tions observed. The possible effects of size-selective
Years 40
20 Nitrogen Availability
kg/ha/year disturbances were indicated by the simulation exer-
0" o
(kg/ha/year) cises. Seed rain does not limit regeneration in the model,
but the distribution of habitats in the study area is
sufficiently coarse-grained that mass effects (Shmida &
Ellner 1984) could be important. Pine is comparatively
Fig. 5. Total basal area (z-axis) of each species in relationto
time and N-status during400 yr of simulation. drought-tolerant, as well as nutrient stress-tolerant, so
the dominance of pine on some of the rocky upland
sites may be partly due to seasonal drought stress. Fi-
nally, casual sowing of tree seeds was sometimes car-
ried out after cutting (Back & Renstrom 1988), particu-
larly in the later 19th century, which may have biased
the species composition of the resulting stands.
In their simplest forms, both gradient analysis and
calibration imply some form of static mapping between
vegetation and site factors. The results of these
modeling exercises indicate several sources of compli-
cations for this traditional mapping. These include the
 - Simulation modeling of a boreal forest landscape - 611

time-dependent nature of the major species responses References

to the fertility gradient and the effect of size-selective
disturbances on inferred age and N-status. In general, Aber, J. D. & Melillo, J. M. 1982. FORTNITE:A computer
the dependence of vegetation on site factors is not model of organic matterand nitrogen dynamics in forest
ecosystems. University of Wisconsin Research Bulletin
functional but relational (Roberts 1987); site factors
R3130.
constrain what is possible but do not determine what
Back, K. & Renstr6m,S. 1988. Vegetationandforest condi-
vegetation actually occurs. tions in the Storesjondistrict, TivedenNational Park (in
The changes in the correlations between the direct Swedish). Swedish EnvironmentalMonitoring Program
species responses to the simulated N-status gradient (PMK), EnvironmentalProtectionBoard, Uppsala.
and the simulated basal area were largely as expected. Beals, E. W. 1984. Bray-Curtisordination:an effective strat-
The interesting feature was the fairly sharp transition egy for analysisof multivariateecological data.Adv.Ecol.
at 100-120 yr. Light competition influences the overall Res. 14: 1-55.
stand composition only after the trees are fairly large, Bonan, G. B. 1989. Environmentalfactors and ecological
processes controllingvegetationpatternsin borealforests.
and the onset of this effect seems to be rapid. I. C.
LandscapeEcology 3:111-130.
Prentice (pers. comm.) pointed out that this may partly
Botkin, D. B. 1981. Causalityand succession. In:West, D. C.,
explain some of the contrasting concerns of pure and Shugart,H. H. & Botkin, D. B. (eds.) Forest Succession.
applied forest ecologists. Foresters are usually con- Springer-Verlag,New York.
cerned with young stands where the growth (and pos- Botkin, D. B., Janak,J. F. & Wallis, J. R. 1972. Some ecologi-
sibly species composition) are more closely determined cal consequencesof a computermodel of forest growth.J.
by the nutrient status. Forest ecologists have been more Ecol. 60: 849-873.
concerned with species replacements that occur over Ellenberg,H. 1971. Nitrogen content mineralizationand cy-
cling. In:Productivityofforest ecosystems.Proc. Brussels
longer periods of time; light competition is crucial in
Symp. 1969, UNESCO.
determining the longer term pattern (Prentice &
Faith,D. P., Minchin,P. R., & Belbin, L. 1987. Compositional
Leemans 1990).
dissimilarityas a robust measure of ecological distance.
Prentice & Leemans (1990) predicted that direc- Vegetatio69: 57-68.
tional succession would be the norm for undisturbed Fulton, M. R. in press a. A computationallyefficient forest
boreal forest, and any other forest in which narrow- successionmodel:design andinitialtests. For. Ecol. Man-
crowned trees predominate. Two results presented here age.
indicate that forest stand structures in this system also Fulton, M. R. in press b. Adult recruitmentas a function of
converge over time: (1) stand simulations affected by juvenile growth rate in size-structuredplant populations.
size-selective disturbances were different from the 'nor- Oikos.
Harrison,E. A. & Shugart,H. H. 1990. Evaluatingperform-
mal' course of succession at first, but later became
ance of an Appalachian oak forest dynamics model.
indistinguishable from undisturbed stands; (2) the in- Vegetatio 86: 1-13.
ferred age of the field plots was positively correlated Helmisaari, H. & Nikolov, N. 1989. Survey of ecological
with the similarity to the matched model stands. Adult characteristicsof boreal tree species in Fennoscandiaand
tree interactions, if allowed to proceed uninterrupted, the USSR.WorkingpaperWP-89-65, InternationalInsti-
increasingly constrain the range of possible structures tute for Applied Systems Analysis, Laxenburg.
and species compositions in this system. Hill, M. 0. & Gauch, H. G. 1980. Detrendedcorrespondence
analysis, an improvedordinationtechnique.Vegetatio42:
47-58.
Jongman,R. H. G., terBraak,C. J. F. & van Tongeren,0. F. R.
Acknowledgements. The authoris indebtedto H. Hyttebom,
1987. Data analysis in communityand landscapeecology.
M. Sykes, and especially I. C. Prenticefor advice andconver-
sations. S. Brakenhielmkindly providedthe field data on the Pudoc, Wageningen.
Lid, J. 1985. Norsk, svensk, finsk flora.Det Norske Samlaget,
study area in Tiveden National Park, and sharedhis knowl-
Oslo.
edge of the area's naturalhistory. I. C. Prentice and three
Mack, R. N. & Harper, J. L. 1977. Interference in dune
anonymousreviewersprovidedhelpfulcommentson themanu-
annuals:spatialpatternandneighborhoodeffects. J. Ecol.
script. This work was supportedby a grantto I. C. Prentice
65: 345-365.
fromtheSwedishNaturalScienceResearchFoundation(NFR).
Minchin,P. R. 1987. An evaluationof the relativerobustness
of techniques for ecological ordination. Vegetatio 67:
1167-1179.
Monsi, M., Uchijima, Z. & Oikawa, T. 1973. Structureof
foliage canopies andphotosynthesis.Ann.Rev. Ecol. Syst.
4: 301-327.
Pastor,J. & Post, W. M. 1985. Developmentof a linkedforest
productivity-soil carbon and nitrogen model. Technical
612 Fulton, M.R.

Manual ORNL/TM-9519. Oak Ridge National Labora- Smith, T. & Huston, M. 1989. A theory of the spatial and
tory, Oak Ridge, Tennessee. temporaldynamics of plant communities. Vegetatio 83:
Prentice, I. C. 1980. Vegetation analysis and orderinvariant 49-69.
gradientmodels. Vegetatio42: 27-34. ter Braak,C. J. F. & Prentice,I. C. 1988. A theoryof gradient
Prentice,I. C. & Leemans, R. 1990. Patternand process and analysis.Adv. Ecol. Res. 18: 271-317.
the dynamics of forest structure:A simulationapproach. Waring,R. H., Schroeder,P. E. & Oren,R. 1982. Application
J. Ecol. 78: 340-355. of the pipe model theoryto predictcanopy leaf area.Can.
Roberts, D. W. 1986. A dynamical systems perspective on J. For. Res. 12: 556-560.
vegetation theory. Vegetatio69: 27-33. Watt,A. S. 1947. Patternandprocess in the plantcommunity.
Shmida, A. & Ellner, S. 1984. Coexistence of plant species J. Ecol. 35: 1-22.
with similarniches. Vegetatio58: 29-55. Weiner,J. & Thomas,S. C. 1986. Size variabilityandcompe-
Shugart,H. H. 1984. A theory of forest dynamics. Springer- tition in plant monocultures.Oikos 47: 211-222.
Verlag, New York.

Received 19 January1991;
Revision received 5 July 1991;
Accepted 9 July 1991.