Forest Succession Models
Author(s): H. H. Shugart, Jr. and D. C. West
Source: BioScience, Vol. 30, No. 5 (May, 1980), pp. 308-313
Published by: Oxford University Press on behalf of the American Institute of Biological
Sciences
Stable URL: http://www.jstor.org/stable/1307854
Accessed: 08-03-2017 14:36 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/1307854?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
American Institute of Biological Sciences, Oxford University Press are collaborating with
JSTOR to digitize, preserve and extend access to BioScience
This content downloaded from 132.174.254.127 on Wed, 08 Mar 2017 14:36:57 UTC
All use subject to http://about.jstor.org/terms
Forest Succession Models
H. H. Shugart, Jr., and D. C. West
the proceedings of two recent workshops
on modeling forest dynamics (Fries 1974
in forestry, Slatyer 1977 in ecology),
applied as tools for studying forest
preoccupied the attention of a large and
porbehavior. Some of the more recent modthere is only one common literature citation of ecologists during the last 100
tion. This is due, in part, to the newness
els are capable of simulating successionyears. Studies in succession attempt to
of the models and, in part, to the pubdetermine the changes in species compo- al patterns and have been used in devellication of many forestry models in what
oping new insights into the nature of
sition and other ecosystem attributes
(e.g., biomass, diversity) expected over long-term forest dynamics and ecologicala nonforester would consider rather obThe study of all natural phenomena in-
son and Christofolini 1966). These and
cluded under the term succession has
other forest simulators have been tested
time. Because successional studies have
succession.
Simulations in the early 1960s were
have system dynamics over long time pe- primarily aimed at production budgets,
riods, the exact patterns of these ecosys- element cycling in plant-soil systems, or
trophic-level dynamics. But exploration
tem dynamics often are inferred rather
of mathematical, theoretical, and comthan directly measured. The subjectivity
been concentrated on forests that tend to
scure places.
In this paper, then, we will review the
models developed in forestry and ecology, with an emphasis on models of ecological succession.
puter applications in ecology had beTYPES OF FOREST SIMULATIONS
come a dominant theme in ecological reThere are three basic organizational
search by the late 1960s (Patten 1971,
categories of forest simulation models
1972, 1974, 1975), and papers in several
mechanisms involved in succession. In(see Table 1):
journals touted the future of systems
ecology as the logical consequence of
deed, Gleason's 1926 statement* Tree models take the individual tree
trends emphasizing quantitative ecology
as
the basic unit of a forest simulator.
American ecologists as well, dis(Davidson and Clymer 1966, Garfinkel
cussing the fundamental nature, strucThe degree of complexity ranges from
1962, Watt 1966). The International Bioture, and classification of plant associsimple tabulation of the probabilities of
logical Programme Biome, for example,
ations, and their chronic inability to
an individual tree of one kind being reused ecosystem models as a central
come to any general agreement on the
placed by an individual of another kind,
theme. Ecologists also had begun to rematters, make it evident that the last
to extremely detailed models that inword is not been said on the subject.
consider the underlying mechanisms include three-dimensional geometry of difvolved in ecological succession (Drury
ferent species at different sizes.
-is as valid today as it was 50 years ago.
and Nesbit 1973, Odum 1969), which
* Gap models dynamically simulate
Recently ecologists interested in
generated further interest in the longparticular attributes of each individual
studying succession began using matheterm dynamics of ecosystems. This intree on a prescribed spatial unit of relamatical models of forest dynamics. Modtellectual climate made the development
tively small size-usually either a gap in
els have the advantage of being formal
of models of ecological succession a logithe forest canopy or a sample quadrat.
descriptions of inferences about succescal and needed step toward progress in
* Forest models consider the forest as
sional mechanisms, which can be anathe analysis of forest ecosystems.
the focal point of the simulation model.
lyzed to produce predictions about longAt the same time, foresters also recogForestry yield tables constitute a highly
term ecosystem dynamics. By the early
nized the potential for computers in their
data-dependent subset of forest models.
1960s, forest biologists at several institufield, and papers addressing the future
tions were independently using digital
importance of forest models appeared in
In general, the model type used is
computers to design mathematical modtrade and forestry research journals. Rebased on the problem under consideraels of changes in forest composition
search foresters realized that yield tation, the data available, and the desire to
(Hool 1966, Odum 1960, Olson 1963, 01bles, the mainstay of prediction of forest
develop a flexible model. The tree and
yields, were not flexible enough to be
forest model categories correspond to
used if the environment changed, if the
the tree and stand model categories used
Shugart is a senior research staff member, and West
is a research staff member-both at the Environmengenetics of the planted trees were imin Munro's (1974) review of forestry
tal Sciences Division, Oak Ridge National Laboratoproved greatly, or if forests were ferti- models. In our review, we recognize gap
ry, Oak Ridge, TN 37830. Research supported by
models (which might be considered as a
lized. So they began exploring modeling
the National Science Foundation's Ecosystem Studies Program under Interagency Agreement No.
techniques to achieve a more mechaspecial case of tree models) as a cateDEB77-25781 with the U.S. Department of Energy
nistic approach to understanding tree
gory
developed exclusively for use in
under contract W-7405-eng-26 with Union Carbide
studying ecological succession. In about
growth and forest yield.
Corporation. Publication No. 1481, Environmental
of the inferred patterns has fueled considerable debate on the general nature of
succession and has led to seemingly interminable confusion on the biological
Sciences Division, ORNL. ?1980 American Institute of Biological Sciences. All rights reserved; the
U.S. government retains a nonexclusive, royaltyfree license to publish or reproduce this article.
Exchange of information between
ecologists and foresters on their forest
models seems limited. For example, in
one-half of the models we have consid-
ered, the authors have included features
that allow simulations of long time
BioScience Vol. 30 No. 5
308
This content downloaded from 132.174.254.127 on Wed, 08 Mar 2017 14:36:57 UTC
All use subject to http://about.jstor.org/terms
periods (e.g., 200 years), in which case
we have categorized them as being
applicable to studying succession.
TABLE 1. Categories of forest dynamic models and examples of each type of model.
Succession models (indicated by rectangles) are able to simulate forest dynamics
over time scales exceeding the life spans of the species considered.
CATEGORY AGE-STRUCTURE DIVERSITY SPACE REFERENCE SUCCESSION MODELS SPECIES OR FOREST TYPE
Tree Models
9Nwnhoi. 1964' Pseudosugo menziei
L..
1967'
Mitchll
Monospecies spatial tree models (two
categories in Table 1)-whether or not
Pinus
1969*
contot
Ptc"o
g9uca
Lin 1970' Tsu1o he-tophyl.
spatial Bela 1971' Popu.6s tremuods
Hatch
H.gyi
1971*
1974
Pnus
Pinu
res-oso
bonksioo.
. Lin 1974 Pseudotsugo menzs Trsug.a heopyl
they consider even-age or mixed-age
I mono0-specis Cl/n. 1903 Pmu, k
stands-are used almost exclusively in
. nonTpotiol Gouldin. 1972' Pseudoto en
sophisticated evaluations of planting,
spacing, and harvesting schemes in commercial forests. The information they
produce is communicated to large governmental or industrial land managers
usually by direct means (e.g., internal reports), rather than through the scientific
Curtis 1967 Psudosugo m.nzis.
Dross 1970'
Sullivan & Clut"r 1972 Pus toedo
*v.-oged Burkhort & Strvb 1974 P,nus toed
Clutt.r
Elfring
probably only a subsample of the ones
actually in use.
These models function by periodically
incrementing individual trees (usually
tree diameter, crown volume, and vari-
ous form and shape parameters)-usually in 1- to 5-year time steps. For example, Mitchell's (1969) model of white
spruce (Picea glauca) uses branch-pruning of trees that overlap to determine
competition interaction.
The models contain equations that explicitly express the crowding of trees and
1974
Pnus
rodioto
Pin.s
sylv't-
TREE ( Adird 1974 P,nus potu., Cupressus . pp
Arn,y 1974 Pseudotsuga 4 menzs,
*spotiol Mitch.ll 1975 Pseudotsugo menzs,
' mono-Sp*cies 1 SPti , 197- S- -p-,,i
No.kon & R.obets 1974 SLoi se.rperv,rens
. mxd-.oged ..Suzuki & U..ura 1974 Chomo.cypr,s spp.
Loak 1970r ' Northorn Hardwood Forest
ForciTr 1975 Northrn Hordwood For.-t
mix.d
scie ( 4"opoCal
1976
Northorn
Hardwood
Fofst
mixed-species
INoble
non,spa,i.,
H_
1976
o.ooS...tiRl
< ---t'" olNobbSldyHorn
? SIyW 19O
Ta*manion
Wot
Sc6rophyll
-Roinfor-s
literature. The models that we have
listed in these categories in Table 1 are
1974
mixed .spcis(on.po.iol
nonspotial
S..... Northern
1974 . .. ..Hordwood Forost
,m.x.d.,p.c...
n 1974
v *po1 io Ek & Mon.rud 1974 Northorn Hordwood For*st
( W.ggon.r & Stoph.ns 1970 Northern Hordood FoIst
GAP mi.xd-og*d mixed.spcis non, p I B.othln et Tl 1972 North.rn Hordwood For.st
Shugort & Wst 1977 Southern Appalachion Foarst
verticol M.ilk. t 1l 1978' Upland Pine-Oak Fo-st
Thorp 1978 Mi,sisippi Floodploin Forst
ShuTlt ? Nobt 190 Montoan Euc.ayptus For,st
< Shugart t l. 1<80. Tropical Roinforest
Hool
1966
North,n
Hardwood
For*st
Olson L Chri.s.olini, 1966 - S..outhrn Appalachi.n For.t
'*:- Mo-sr & Hall 1969 Nor,thrn Hardwood Forst
.mixed.g.d 9mix.ed .specis nonspatil Sh.g,ort *t o 1973 For.t.. in Upp.r Michigon
FOREST
Johnson
&
Shorp*1976
For*sts
in
Georgi
Piedmont
Wilkins 1977' Forsts in Tsmaonia
*ven-oged mono-species nonspatial ( Mo. yi.ld tobl ud..
in aorestry today
* Dissorttion
require the degree of exactness that
tives to monospecies spatial tree models.
some of the models considered are caUsed almost exclusively in pine (Pinus
sp.) plantations, nonspatial models are
pable of providing, and his derivations,
usually in the form of differential equa- based on a typical-size tree in a mixed
tions with basal area, stocking density,
forest as a modeling unit, are probably
reasonable. Suzuki and Umemura (1974)
and volume (biomass) of a forest stand
have attempted to include mixed-age efchanging with respect to time. Since
can be easily adapted to either even- or
fects in nonspatial tree models by solving
mixed-age stands. In fact, Hegyi's (1974) these relationships are functions of the
for shape parameters of the underlying
mixed-age model, and Mitchell's (1969, size of the average tree, the models contain parameters derived from the exstatistical frequency distribution of the
1975) models are modified in the converse manner. Designed for commercial
pected growth of trees. The even-age,diameters of all the trees in a modeled
monospecies character of the simulated
forestry operations, the models do not
stand as dynamic variables.
forests allows the assumption that mathinclude phenomena that ecologists
Mixed-age, mixed- or monospecies,
would expect in a succession simulator.
ematical functions for the expected renonspatial tree models simulate ecologiThey generally ignore establishment of
cal success in naturally regenerated forsponse of an average or typical tree are
invading seedlings and often use funcsufficient to express these relationships
ests. They emphasize birth-death proamong volume, stocking, and basal area.
tions for geometry of trees that could oncesses affecting individual trees and
These models work best if the trees
ly be expected to hold in young, vigorgreatly deemphasize the importance of
tend to be the same size, which explains
ously growing trees. The models
tree growth and form. Because they are
sometimes use thinning or harvest as a
their use in the genetically optimized,
not particularly complex (i.e., birth and
surrogate for mortality.
short-rotation, crop-like Pinus plantadeath of trees might be treated as simple
Because of the level of detail needed,
tions. The underlying assumptions of the stochastic processes; replacement of
these models synthesize great amounts
models limit their application to eventrees as a first-order Markov process),
of autecological data that are usually onage stands, and the development of
the authors frequently attempt to capture
ly available for commercial species, so it
mixed-age models using this approach is
the salient aspects of succession with a
is difficult to extend the models to mixeddifficult. Unlike spatial monospecies
minimal model representation. In this
models, even-age, monospecies nonspecies forests. There is a distinct bias
objective, the models are actually explotoward using this modeling approach
spatial tree models can be solved analytirations into the consequences of theories
with tree species that are commercially
cally in some cases and in all cases reand assumptions on the nature of ecological succession based on the attributes of
important in the Pacific Northwest of the quire only a moderate amount of
United States and in British Columbia.
computer time.
the species involved (Drury and Nesbit
Many of the papers cite C. S. Holling's
Solomon's (1974) even-age, mixed1973, Gleason 1926).
These models can provide considspecies, nonspatial tree model is related
works (e.g., Holling 1966, 1971) as a
source of inspiration in terms of the apto the present category and uses a typical erable insight into patterns of ecosystem
proaches used.
tree at different ages in a system of dydynamics and can be solved analytically
without resorting to digital computation.
namic equations solving for forest-level
Even-age, monospecies, nonspatial
attributes of northern hardwood forest
tree models, also useful in commercial
The Noble and Slatyer (1980) model, for
stands. Solomon's application does not
example, uses the vital attributes of speforestry, are logical nonspatial alterna309
May 1980
This content downloaded from 132.174.254.127 on Wed, 08 Mar 2017 14:36:57 UTC
All use subject to http://about.jstor.org/terms
cies to determine the expected patterns
of community successions generated by
competition among the species. Vital at-
cal, not the horizontal, dimension-a
simplification that greatly reduces the
cost in running them but also eliminates
servations. Generally, logistic considerations in collecting samples limit the range
of conditions over which the models can
consideration of complex spatial patbe tested (e.g., it is difficult to collect data
species uses to persist at a site, the
terns of trees (should this be important inthat relate to assessing a change in climodes for establishment, the availability
a given application). The vertical gap
mate). Many of the forestry-oriented
of a method for persistence (e.g., seeds,
models are probably best used in studies
models (Table 1) are used with specific
vegetative sprouts) at different life stages of successional dynamics of natural forprescribed conditions and can be tested
of the plants (juvenile, mature, propests considered over long time spans.
directly with independent data sets.
agule, extinct), and longevity of individIn most applications, models are deuals. Using these species attributes, they Forest Models
signed to save costs (via reduction in the
construct schematic diagrams of changes
need for large and expensive data collecYield tables used in forestry managethat can be compared with observational
tion programs); so the need to collect
ment are, in fact, empirical models of exdata from a given area.
data to test a model must be optimized
pected responses of an even-age forest
Mixed-age, mixed-species, spatial tree
with the cost of collecting the data. Most
usually of a single species. In this conmodels have only one representative in
models used in studies of succession
text, a forest is taken at a larger spatial
Table 1: FOREST (Ek and Monserud
(Table 1) make direct comparisons bedimension than either single tree or gap
1974). The mathematical functions in the
tween model responses and field obsermodels consider explicitly.
model include form, seed supply, and exvations, but the logistics of collecting
Comparable succession models have
act location of each simulated tree above
field data are a greater problem than in
been developed through a variety of
a certain diameter and height. The model
the short-term forestry applications.
mathematical approaches. Most considcan simulate a forest of any size, because
Logical procedures appeal to underer the landscape to be composed of a
the only limitation is the amount of data
lying principles or model assumptions
number of mosaic elements that change
storage space available on the computer.
that, if true, would be taken as a logical
in response to successional processes.
FOREST is the most complex and debasis to expect the model's predictions
These changes may be viewed as probtailed model of forest succession we
to be correct. For example, Horn (1976)
ablistic (e.g., Hool 1966, Wilkins 1977)
have considered. Because of its level of
assumed that probabilities of species
or as continuous, depending on modeling
detail, the model requires many parametransitions
in succession are ergodic
assumptions relating to the actual size of
ters (e.g., form equations, canopy-shape
(constant over space and time); our own
the landscape considered (Shugart et al.
functions, density and pattern of seed
set of assumptions (Shugart et al. 1973)
1973).
tributes considered are the modes that a
rain associated with different individual
allowed succession to be modeled as a
Gap models simulate succession by
TESTING SUCCESSION MODELS
calculating the competitive interrelations
among trees in a restricted spatial unit- The process of testing a model to
either a gap created by the death and redetermine how much credibility one
moval of a canopy tree or a sample quadshould place in its predictions has been
termed model "validation." This is an
rat. A nonspatial example (Waggoner
and Stephens 1970) uses the probabilityunfortunate term, because testing does
that a forest inventory quadrat, classifiednot make models "valid"; it simply
as one type of forest, will at some
gives one an idea about the reliability
later time have changed to the extent
and, hence, the utility of a given model
that it can be classified as some other
(Mankin et al. 1977). Succession models
which agrees with extant data on forest
Forest models tend to be dependent on
trees). Developed for the forests of Wisdata on rates of change of the mosaic ele- set of ordinary linear differential equaconsin, it would require considerable eftions. Almost any model develops from a
ments assumed to comprise the forests,
fort to modify it for another forest syslogical set of assumptions as an underand the actual mechanisms that cause
tem. The model produces output and
lying basis, and more theoretical studies
the changes in the forests do not appear
predictions at a level of complexity unof succession may use logical procedures
explicitly in the models. All of the forest
equaled in most forest inventories.
as a primary means of anticipating the
models listed in Table 1 require little
applicability of model results.
computer time and can be solved analytiLong-term projections constitute a
Gap Models
cally in many cases.
type of forest. This model is of similar
ters to determine tree height, and then
dynamics. Since most of the vegetation
considered by the models has been disturbed by human activities, this test uses
the model to project the pattern of vegetation after a long period of time. The resultant predictions can then be tested for
consistency with what is known-either
from old records or from relict sites-
about the undisturbed vegetation. Horn
(1976) used such a test in comparing the
are particularly challenging to test bespecies composition at equilibrium precause the phenomena they predict may dicted by his Markovian simulator with
form mathematically to that of Horn
(1976), except that it is applied to a
take years (or even generations) to obchange in stands and not to single-treeserve in nature. Further, the stochastic
replacements.
nature of many models requires large
Other gap models simulate year-to- sample sizes in the test data sets. Thus,
year changes in the diameter of each
tree. They do not account for the exact
location of each tree but use tree diame-
natural test on a succession model,
most succession model tests have to
rely on inference rather than direct
observation.
that of a climax beech forest in the
Princeton Woods. Of the models shown
in Table 1, only those indicated as succession models could reasonably be expected to satisfy this sort of test.
In many instances, the long-term predictions of forest succession models can
Model testing procedures take a varie-
be used to develop a theoretical under-
use simulated leaf area profiles to devise
ty of forms:
competition relationships due to shad"Brute force" procedures compare
ing. These models are spatial in the verti- model responses directly with field ob-
be inspected as new data are collected.
standing of forest dynamics, which can
For example, Ranney et al. (1978) simu-
310
BioScience Vol. 30 No. 5
This content downloaded from 132.174.254.127 on Wed, 08 Mar 2017 14:36:57 UTC
All use subject to http://about.jstor.org/terms
.- A. HOP HORNBEAM
8 100
o
HICI
I
1--
--
ORNL-DWG 78-4871 responses is in the structure (e.g., diame-
KORY- WHITE OAK~__
^ORY WlITE = ter distribution) of the forests. One obvi-
-------
ous advantage of the use of models in
ASH
J 80
--- -- succession research is in exploring such
A. BASSWOOD
- cases, which are not open to direct experimentation with well-tested simula-
a 60
()
m 40
- N. RED OAK B. CHERRY
tion models.
LU
_ _ SUGAR MAPLE
, 20
_J
LU
I
I
^ 0
0
20
I
40
60
I
I
80
I
100
ONLY Hindcast procedures involve using
I
past events as if they were natural
140 160 180 200 iments to test the model's a
120
YEARS FROM 1976
A -HOP
.-- HICKORY
81oo
100
LU 80
BA
60
N.
-
m 40
RED
B.
OAK
I
20
40
iHItE, O A perturbation to test our model on its abil-
I
60
"
\SSWOOD forests not subjected to the chestnut
'--^s^ --blight, based on historical data. Othe
ENTIRE ISLAND _ events that might be used for such h
PROPAGULE INPUT cast tests include other diseases, cl
changes expressed in pollen records, the
SUGAR MAPLE
, 20
0
-
CHERRY
BEECH
ur',
r
gu U
HORNBEAN
HITE OAKity to predict the species composition of
ASH
o)
dict them. For example, we (Shuga
ORNL-DWG 78-4872 West 1977) used the chest
I
80
I
I
100
I
120
I I 2 advent
Europeans,
introduction or
160 of
180
200 or
.
.
140
extinction
YEARS FROM 4976
ds
of
1
to
2
ha
Figure 1. Top: Simulated edge development for forest islan
pected composition of the forest edge for 200 years from rn ieasurements made in 1976 casionally, while the computer program
(year 0 on graph). Bottom: Simulated composition for entire iforest island including edge of a model is being tested or while a
and interior. These simulations are based on the FOREST mo
and are taken from Ranney et al. (1978).
of
sh
del (Ek and Monserud 1974) documented model is being used, the
model may predict a correct response in
lated the expected dynamics of forest This
is- is partic] ularly true in instances in the face of a program error. For exwhich the patiterns predicted are a prod- ample, a model that we developed for the
lands-namely, small relic patches of
forest in a matrix of farm or urban land
uct of higher o
)rder
uses-in Wisconsin using Ek and Mon-
teractions and
different sizes. Further, because the
ciduous/coni
(e.g., competitive) in- southern Appalachian Forest (Shugart
not simply a consequence and West 1977) was predicting forests
of the physiological ranges of the species that, to the best of our knowledge,
serud's (1974) FOREST model. They
considered. F or example, Botkin et al. should occur in the Georgia Piedmont.
predicted the expected equilibrium com(1972)
position and structure of forest islands
of predict ed the location of the de- Through a keypunch error, we had acci-
ferous forest transition; dentally given the model a warmer cli-
her model of a flood- mate than that appropriate for East TenTharp (1978):ested
t
plain forest v vith a similar test using a nessee-a climate that was, in fact,
location and spatial interactions of each
gradient of flood frequency; Shugart and appropriate for the Georgia Piedmont.
simulated tree, they could predict differences in different parts of each forest Noble (1980) i used a combination of alti- Observations of model behavior such
FOREST model explicitly considers the
island. Figure 1 shows a 200-year projec- tude changes and different wildfire fre- as this do not constitute model experiquencies to test their model's ability to
ORNL-DWG 76-8440
100
tion starting from actual data collected in simulate gradients in the vegetation in
1976 and contrasts the overall island
the Brindabella Range in the Australian
tion of a forest island's floristic composi-
80
Alps.
composition with that of the island's
edge. Certain species (e.g., northern red In a model experiment inspecting the
oak) disappear from the edge with time, forest response to a slowly changing cli- ,, 60
and the compositions of edges and interi-mate, we (Shugart et al. 1980) noted a
z 40
ors of the forests are clearly different. hysteretic response in the composition of
02
the
simulated
forest
(Figure
2).
In
this
Since forest islands are to a great desimulation, the composition of forests at 20
gree a consequence of man's land use,
particular sites could be expected to difthey are a new type of natural system for
which we have a limited data base enfer depending upon whether the region
3750 4000 4250 4500 4750 5000 5250 5500
had been under warming or cooling cliabling us to infer the long-term system
DEGREE- DAYS
matic conditions over the previous thoubehavior. Models represent a valuable
Figure 2. Percentage of total stand biosands of years. Such hysteretic readjunct to studies on the fundamental
mass attributable to yellow-poplar (Liriodendron tulipifera) under gradually changsponses have been noted for years,
nature of such ecosystems.
ing, growing degree day values. One curve
manifested as differences in the locations
Predicting gradient responses in(dashed line) illustrates the mean value of
of community boundaries in mountains
volves running the model under a set of
5 simulated 1/12 ha forest stands as the anor following disturbances (Griggs 1946,
differing conditions that approximate
nual value of growing degree days is inMarie-Victorin 1929, Polunin 1937), but
creased from 3800 to 5300 at a rate of 1.0
some naturally occurring ecological graz
tr
dient. If the model can predict patterns
of vegetation along this gradient, it has
then passed a rather severe test of the
range of conditions over which it can be
expected to produce reasonable results.
the actual mechanisms involved in such
degree day year-1 following an initial equi-
librium period of several hundred years.
responses are not amenable to direct exSolid line illustrates the same conditions
perimentation because of the long timewith growing degree days reduced from
scales involved. With model experi5300 to 3800 growing degree days. (From
Shugart et al. 1980b)
ments, one possible mechanism for such
311
May 1980
This content downloaded from 132.174.254.127 on Wed, 08 Mar 2017 14:36:57 UTC
All use subject to http://about.jstor.org/terms
QUERCUS VELUTINA
ments and rarely surface in the professional literature, but they do provide the
user with a sense of the reliability of the
75 -
70 -
model's predictions. We have learned
from various personal communications
that others, who have either developed
or used forest succession models, have
experienced these difficult-to-derivefrom-intuition anecdotal occurrences.
Their importance as exploratory model
tests is probably underestimated.
bia, Vancouver.
Bosch, C. A. 1971. Redwoods: a population
65 -
model. Science 162: 345-349.
60 -
Botkin, D. B. 1973. Estimating the effects of
55 -
carbon fertilization on forest composition
, 50 -
by ecosystem simulation. Pages 328-344 in
, 45 I
- 40 -
G. M. Woodwell and E. V. Pecan, eds.
35-
Carbon and the Biosphere. Proceedings,
50
24th Brookhaven Symposium in Biology.
30 -
CONF-720510. National Technical Infor-
25 -
MODEL APPLICATIONS
mation Service, Springfield, VA.
. 1977. Forests, lakes, and the an-
20 -
thropogenic production of carbon dioxide.
15 -
Most models built strictly for forestry
use are usually intended for applications
in a restricted set of specified circum-
BioScience 27: 325-331.
10 -
5 -
stances. However, several of the succes-
sion models presented in Table 1 have
been used to evaluate environmental im-
Bella, I. E. 1971. Simulation of Growth, Yield
and Management of Aspen. Unpublished
Ph.D. Thesis, University of British Colum-
0 50 400 150 200 250 300 350 400 450 500
YEARS
Figure 3. Changes in biomass in black oak
Botkin, D. B., J. F. Janak, and J. R. Wallis.
1972. Some ecological consequences of a
computer model of forest growth. J. Ecol.
60: 849-872.
Burkhart, H. E., and M. R. Strub. 1974. A
(Quercus velutina) starting from bare
pacts on naturally occurring forests. Botmodel for the simulation of planted loblolly
ground at year 0 with a 10% reduction in
kin (1973, 1977) considered the effects ofgrowth rate compared to control (black
pine stands. Pages 128-135 in J. Fries, ed.
line). When the species is stressed from
CO2 enrichment on plant growth and
Growth Models for Tree and Stand Simulasubsequent effects on forest dynamics. year 0, there is an appreciable reduction in tion. Res. Notes 30, Department of Forest
biomass through succession. Stress apYield Research, Royal College of Forestry,
He found that an arbitrarily assumed
plied at year 50 was an appreciable effect
Stockholm.
percentage change in rate of photosynonly after 100 years, and stress applied at
year 400 had no effect. These very different Clutter, J. L. 1963. Compatible growth and
thate production at the individual plant
yield models for loblolly pine. For. Sci. 9:
responses in the reaction of black oak to
level in C02-enriched atmospheres was
354-371.
altered growth are a product of a complex
not manifested directly as a change in
forest increment. Other effects, such as
plant competition and shading, tended
to lower the magnitude of the system
interaction of forest structure and the vigor
of the trees stressed. (From West et al.
1980)
. 1974. A growth and yield model for
Pinus radiata in New Zealand. Pages 136-
160 in J. Fries, ed. Growth Models for Tree
and Stand Simulation. Res. Notes 30, De-
response.
partment of Forest Yield Research, Royal
McLaughlin et al. (1978) and West et
al. (1980) performed model experiments
Important future applications of sucCollege of Forestry, Stockholm.
cession models, particularly gap models,
Curtis, R. 0. 1967. A method of estimation of
(Figure 3) on chronic air pollution stress,would involve evaluating large-scale and
gross yield of Douglas-fir. For. Sci.
Monogr. 13: 1-24.
expressed as a change in growth rates of long-term changes in the ambient levels
pollution-sensitive trees. They noted
of pollutants and assessing the effects of Davidson, R. S., and A. B. Clymer. 1966. Desirability and applicability of simulating
that the response of growth over the
climate change. If human activities alter
ecosystems. Ann. N. Y. Acad. Sci. 128:
long-term in natural forests might varyenvironmental
in
conditions on a global
790-794.
direction as well as in magnitude from
scale, models will become increasingly
Dress, P. E. 1970. A System for the Stochaswhat one might predict from laboratory important as tools for prediction. This
tic Simulation of Even-Aged Forest Stands
or greenhouse studies.
has been true of almost any application
of Pure Species Composition. Unpublished
All of these studies identify a common in which ecologists made observations
Ph.D. Thesis, Purdue University, West Laproblem-namely, that in natural forests
on the behavior of forest ecosystems in
fayette, IN.
an environment that had been altered in
in which trees vary in spacing, size, and
Drury, W. H., and I. C. T. Nesbit. 1973. Succompetitive responses, one cannot exsome way. At some level, this case is
cession. J. Arnold Arbor. Harv. Univ. 54:
trapolate directly from laboratory studprobably already appropriate to all forest331-368.
ies to field conditions. Forest succession
models can provide a necessary adjunct
to laboratory-based assessments of environmental effects.
Johnson and Sharpe (1976) have used
their model to inspect land-use changes
ecosystems.
of the structure in unthinned stands of
REFERENCES CITED
Adlard, P. G. 1974. Development of an empirical competition model for individual trees
within a stand. Pages 22-37 in J. Fries, ed.
Growth Models for Tree and Stand Simulation. Res. Notes 30, Department of Forest
Yield Research, Royal College of Forestry,
in the Georgia Piedmont in response to
different levels of forest harvesting and
fire. We suspect that most of the forest
models we have listed could (with proper
Stockholm.
changes in parameters) be used toward a
Amey, J. D. 1974. An individual tree model
similar goal. Such applications might
motivate regional planners to take great-
Elfving, B. 1974. A model for the description
for stand simulation in Douglas-fir. Pages
38-46 in J. Fries, ed. Growth Models for
er consideration of the dynamics of land- Tree and Stand Simulation. Res. Notes 30,
scapes as opposed to the more static apDepartment of Forest Yield Research, Royal College of Forestry, Stockholm.
proach in use today.
Scots pine. Pages 161-179 in J. Fries, ed.
Growth Models for Tree and Stand Simulation. Res. Notes 30, Department of Forest
Yield Research, Royal College of Forestry,
Stockholm.
Ek, A. R., and R. A. Monserud. 1974. FOREST: A Computer Model for Simulating the
Growth and Reproduction of Mixed Forest
Stands. Research Report A2635, College of
Agricultural and Life Sciences, University
of Wisconsin, Madison. Pages 1-14 + 3 Appendices.
Forcier, L. K. 1975. Reproductive strategies
and the co-occurrence of climax tree species. Science 189: 808-809.
BioScience Vol. 30 No. 5
312
This content downloaded from 132.174.254.127 on Wed, 08 Mar 2017 14:36:57 UTC
All use subject to http://about.jstor.org/terms
Fries, J., ed. 1974. Growth Models for Tree
and Stand Simulation. Res. Notes 30, Department of Forest Yield Research, Royal
College of Forestry, Stockholm.
71 in G. S. Innis, ed. New Directions in the
Analysis of Ecological Systems. Part I.
Simulation Council of America, LaJolla,
CA.
ture, Composition and Importance to Regional Forest Dynamics. EDFB/IBP-78/1.
Oak Ridge National Laboratory, Oak
Ridge, TN.
Garfinkel, D. 1962. Digital computer simula- Marie-Victorin, F. 1929. Le dynamisme dans
Shugart, H. H., T. R. Crow, and J. M. Hett.
tion of ecological systems. Nature 194:
la flore du Quebec. Contrib. Inst. Bot.
1973. Forest succession models: a rationale
856-857.
Univ. Montreal No. 13, p. 77ff.
and methodology for modeling forest sucGleason, H. A. 1926. The individualistic conMcLaughlin, S. B., D. C. West, H. H. Shucession over large regions. For. Sci. 19:
cept of the plant association. Bull. Torrey gart, and D. S. Shriner. 1978. Air pollution
203-212.
Bot. Club 53: 7-26.
effects on forest growth and succession: ap- Shugart, H. H., and D. C. West. 1977. DevelGoulding, C. J. 1972. Simulation Techniques
plications of a mathematical model. Pages
opment of an Appalachian deciduous forest
for a Stochastic Model or the Growth of
1-16 in H. B. H. Cooper, Jr., ed. Prosuccession model and its application to asDouglas Fir. Unpublished Ph.D. Thesis,
ceedings, 71st Meeting of the Air Pollution sessment of the impact of the chestnut
University of British Columbia, VanControl Association. Document No. 78blight. J. Environ. Manage. 5: 161-179.
couver.
24.5. APCA, Houston, TX.
Shugart, H. H., A. T. Mortlock, M. S. HopGriggs, R. F. 1946. The timberlinesMielke,
of North
D. L., H. H. Shugart, and D. C.
kins, and I. P. Burgess. 1980a. A Computer
America and their interpretation. Ecology
West. 1978. A Stand Model for Upland For- Simulation Model of Ecological Succession
27: 275-289.
ests of Southern Arkansas. ORNL/TMin Australian Subtropical Rainforest.
Hatch, C. R. 1971. Simulation of an Even6225. Oak Ridge National Laboratory, Oak ORNL/TM-7029. Oak Ridge National LabAged Red Pine Stand in Northern MinneRidge, TN.
oratory, Oak Ridge, TN.
sota. Unpublished Ph.D. Thesis, UniversiMitchell, K. J. 1969. Simulation of growth of
Shugart, H. H., and I. R. Noble. 1980. A comty of Minnesota, St. Paul.
even-aged stands of white spruce. Yale
puter model of succession and fire response
Univ. Sch. For. Bull. 75: 1-48.
Hegyi, F. 1974. A simulation model for manof the high altitude Eucalyptus forests of
aging jack-pine stands. Pages 74-87 in J.
. 1975. Dynamics and simulated
the Brindabella Range, Australian Capital
Fries, ed. Growth Models for Tree and
yield of Douglas-fir. For. Sci. Monogr. 17:Territory. Austr. J. Ecol., in press.
Stand Simulation. Res. Notes 30, Depart- 1-39.
Shugart, H. H., W. R. Emanuel, and D. C.
ment of Forest Yield Research, Royal Col- Moser, J. W., and 0. F. Hall. 1969. Deriving
West. 1980b. Environmental gradients in a
lege of Forestry, Stockholm.
growth and yield functions for uneven-agedbeech-yellow poplar stand simulation modHolling, C. S. 1966. The strategy of building forest stands. For. Sci. 15: 183-188.
el. Math. Bios., in press.
Munro, D. D. 1974. Forest growth models-a Slatyer, R. 0. 1977. Dynamic Changes in Termodels of complex biological systems.
prognosis. Pages 7-21 in J. Fries, ed.
Pages 195-214 in K. E. F. Watt, ed. Sysrestrial Ecosystems: Patterns of Change,
tems Analysis in Ecology. Academic Press, Growth Models for Tree and Stand Simula-Techniques for Study and Applications to
New York.
tion. Res. Notes 30, Department of Forest Management. MAB Technical Notes 4,
Yield Research, Royal College of Forestry,UNESCO, Paris.
. 1971. Background paper for resource simulation games: GIRLS: Gulf Islands Recreational Land Simulation. Resource Science Centre, University of
British Columbia, Vancouver.
Hool, J. N. 1966. A dynamic programingMarkov chain approach to forest production control. For. Sci. Monogr. 12: 1-26.
Horn, H. S. 1976. Succession. Pages 187-204
in R. M. May, ed. Theoretical Ecology:
Principles and Applications. Blackwell Scientific Publishers, Oxford.
Johnson, W. C., and D. M. Sharpe. 1976. An
analysis of forest dynamics in the northern
Georgia Piedmont. For. Sci. 22: 307-322.
Stockholm.
Solomon, D. S. 1974. Simulation of the devel-
Namkoong, G., and J. H. Roberts. 1974. Ex- opment of natural and silviculturally treated
tinction probabilities and the changing age stands of even-aged Northern Hardwoods.
structure of redwood forests. Am. Nat. 108:
Pages 327-352 in J. Fries, ed. Growth Mod355-368.
els for Trees and Stand Simulation. Res.
Newnham, R. M. 1964. The Development of aNotes 30, Department of Forest Yield ReStand Model for Douglas-Fir. Unpublishedsearch, Royal College of Forestry, StockPh.D. Thesis, University of British Colum-holm.
bia, Vancouver.
Sullivan, A. D., and J. L. Clutter. 1972. A siNoble, I. R., and R. 0. Slatyer. 1980. The efmultaneous growth and yield model for lobfect of disturbance on plant succession.
lolly pine. For. Sci. 18: 76-86.
Proc. Ecol. Soc. Aust. 10, in press.
Suzuki, T., and T. Umemura. 1974. Forest
Odum, H. T. 1960. Ecological potential and
transition as a stochastic process II. Pages
analogue circuits for the ecosystem. Am.
358-379 in J. Fries, ed. Growth Models for
Sci. 48: 1-8.
Tree and Stand Simulation. Res. Notes 30,
Odum, E. P. 1969. The strategy of ecosystem
Department of Forest Yield Research, Roynorthern hardwoods predicted by birth and
development. Science 164: 262-270.
al College of Forestry, Stockholm.
death simulation. Ecology 51: 794-801.
Leak, W. B. 1970. Successional change in
Olson, J. S. 1963. Analog computer models Tharp, M. L. 1978. Modeling Major Perfor movement of isotopes through ecosys- turbations on a Forest Ecosystem. Unpubtems. Pages 121-125 in V. Schultz and A.
lished M.S. Thesis, University of Tennes43: 387-388.
W. Klements, eds. Proceedings First Na- see, Knoxville.
Lin, J. Y. 1970. Growing Space Index and
tional Symposium on Radioecology. Rein-Waggoner, P. E., and G. R. Stephens. 1970.
Stand Simulation of Young Western Hemhold, New York, and AIBS, Washington, Transition probabilities for a forest. Nature
lock in Oregon. Unpublished Ph.D. Thesis,
DC.
225: 1160-1161.
Lee, Y. 1967. Stand models for lodgepole pine
and limits to their application. For. Chron.
Duke University, Durham, NC.
Olson, J. S., and G. Christofolini. 1966. ModWatt, K. E. F. 1966. The nature of systems
el simulation of Oak Ridge vegetation suc- analysis. Pages 1-14 in K. E. F. Watt, ed.
models for Douglas-fir and western hemcession. Pages 106-107 in ORNL/TM-4007.
Systems Analysis in Ecology. Academic
lock in the Northwestern United States.
Oak Ridge National Laboratory, Oak
Press, New York.
Pages 102-118 in J. Fries, ed. Growth ModRidge, TN.
West, D. C., S. B. McLaughlin, and H. H.
els for Tree and Stand Simulation. Res.
Patten, B. C. 1971, 1972, 1974, 1975. SystemsShugart. 1980. Simulated forest response to
Notes 30, Department of Forest Yield ReAnalysis and Simulation in Ecology. Vols.chronic air pollution stress. J. Environ.
search, Royal College of Forestry, Stock- 1, 2, 5, and 4. Academic Press, New York. Qual. 9: 43-49.
. 1974. Stand growth simulation
holm.
Polunin, N. 1937. The birch forests of GreenWilkins, C. W. 1977. A Stochastic Analysis of
Mankin, J. B., R. V. O'Neill, H. H. Shugart, land. Nature 140: 939-940.
and B. W. Rust. 1977. The importance ofRanney, J., W. C. Johnson, and H. H. Shuvalidation in ecosystem analysis. Pages 63- gart. 1978. Edges of Forest Islands: Struc-
the Effect of Fire on Remote Vegetation.
Unpublished Ph.D. Thesis, University of
Adelaide, South Australia.
May 1980
313
This content downloaded from 132.174.254.127 on Wed, 08 Mar 2017 14:36:57 UTC
All use subject to http://about.jstor.org/terms