Eur J Forest Res (2005) 124: 133–142
DOI 10.1007/s10342-005-0061-y
O R I GI N A L P A P E R
J. J. Corral Rivas Æ J. G. Álvarez. González
Oscar Aguirre Æ F. J. Hernández
The effect of competition on individual tree basal area growth
in mature stands of Pinus cooperi Blanco in Durango (Mexico)
Received: 22 July 2004 / Accepted: 28 February 2005 / Published online: 13 May 2005
Springer-Verlag 2005
Abstract In this paper, we evaluated how well-selected
distance-dependent and distance-independent competition indices explain individual tree basal area growth of
trees, growing in mature and even-aged stands of Pinus
cooperi Blanco. A total of 18 competition measures were
analyzed of which six do not need tree location (distance-independent) and 12 that utilize tree location
(distance-dependent). The competition situation of a
stand and/or an individual tree was studied using 11
different competitor selection methods. The mean square
error reduction relative to no-competition was used to
judge the performance of each competition index. It was
found that the best distance-independent competition
indices performed as well as the best distance-dependent
competition indices. It was concluded that the BALMOD-index would be a good competition index to be
incorporated into further individual tree basal area
growth models for the study area.
Keywords Pinus cooperi Æ Competition indices Æ
Individual basal area growth
J. J. C. Rivas (&)
Institut für Waldinventur und Waldwachstum,
Georg-August-Universität Göttingen, Büsgenweg 5,
37077 Göttingen, Germany
E-mail: jcorral@gwdg.de
Tel.: +49-551-393554
Fax: +49-551-399787
J. G. Á. González
Departamento de Enxeñerı́a Agroforestal,
Escola Politécnica Superior de Lugo,
Universidade de Santiago de Compostela,
Campus Universitario s/n, 27002 Lugo, Spain
O. Aguirre
Facultad de Ciencias Forestales,
Universidad Autónoma de Nuevo León,
67700 Linares, NL, Mexico
F. J. Hernández
Instituto Tecnológico Forestal, Mesa del Tecnológico s/n,
34950 El Salto, P.N., Durango, Mexico
Introduction
Competition among trees implies that resource supplies
are below levels capable of supporting optimal growth of
two or more trees (Brand and Magnussen 1988; Holmes
and Reed 1991; Gadow and Hui 1999; Pretzsch 2002).
Tomé and Burkhart (1989) and Preztsch (2001) have
shown that individual tree growth models can be used to
analyze how individual tree basal area growth varies
with different levels of competition. The effects of competitor trees on the growth of individual trees have long
been studied by numerous authors in an attempt to
predict growth of different species as accurately as possible (e.g. Clark and Evans 1954; Alemdag 1978; Lorimer 1983; Tomé and Burkhart 1989; Biging and
Dobbertin 1995; Bachmann 1998; Alvarez et al. 2003).
Competition indices may be distance-independent or
distance-dependent. The distance-independent ones do
not require tree coordinates since they are simple functions of stand-level variables or of the initial dimensions
of the subject tree, while distance-dependent models require the dimensions and the relative locations of several
neighbors for the computation. Distance-independent
indices are easy to calculate and less demanding in data
field measurements, which is an advantage.
In Mexico, only few investigations on competition
among trees for resources have been reported (Rodrı́guez 1987; Romero 1993; Valles et al. 1998; Torres
2000). Research carried out in other countries has shown
improvements in diameter growth prediction for different species when a competition index was included (e.g.
Bella 1971; Pukkala and Kolström 1987; Tomé and
Burkhart 1989; Biging and Dobbertin 1992, 1995; Pretzsch et al. 2002; Alvarez et al. 2003). In these studies, a
single tree competition index could not be shown to be
superior, but some indices seemed to perform better with
certain species depending on specific situations. Thus, it
seems reasonable to examine the possibilities of using
competition indices in growth models under Mexican
forest conditions.
134
The objectives of this work are: (a) to compare the
predictive capability of selected distance-independent
competition indices against distance-dependent ones and
(b) to identify the best competition index and/or best
combination of a competition index with a method for
selecting competitors. The selection method should not
require too many field measurements and should be
easily incorporated into an individual tree basal area
growth model for Pinus cooperi in the forest region of El
Salto, Durango, Mexico.
Materials and methods
Data
In the Mexican forests, P. cooperi Blanco is one of the
most important commercial timber species, due to its
good wood properties, wide distribution range and the
harvested timber volume. P. cooperi occurs mainly in the
Sierra Madre Occidental in southern Chihuahua, eastern
Sinaloa, and Durango, at elevations of 1,500–2,800 m.
The species may form pure stands, but usually is mixed
with other pines, junipers, or various species of Quercus
(Garcia and Gonzalez 1998). The stands are moderately
managed with an average cutting cycle of 10–15 years.
Data from eight permanent plots located in mature
and even-aged stands of P. cooperi Blanco, each covering an area of 0.5 ha, were used in this study. The stands
originated from natural regeneration and are approximately 60-years-old. The plots were established in January 1998 and have been remeasured each year as part
of a long-term experiment designed to evaluate growth
and survival in the forest Ejido1 known as La Victoria, in
the province of El Salto, Durango (Mexico). The study
area is located in the Sierra Madre Occidental, between
2340¢–2447¢ latitude and 10521¢–10529¢ longitude,
100 km to the southwest of Durango City (Fig. 1).
The following measurements were taken for each tree
within the plots: X and Y coordinates (m), total height
(m), diameter at breast height (cm), average crown
diameter at the base of the live crown (m) and crown
length (m). In addition, the variables basal area, stem
number per hectare, site index at a reference age of
50 years using the equation proposed by Corral et al.
(2004), and relative spacing index were calculated for
each plot. The relative spacing index is a stand density
measure, which is expressed in this work by the ratio of
the average distance between the trees growing in ploti
(m) and the dominant ploti-height (m). For computing
the effect of competition on individual tree growth, a 10m buffer zone was used around the plot perimeter, the
effective plot size was thus reduced from 100·50 m to
90·40 m. A growth period of 5 years was used in this
study because the annual diameter increment was not
Fig. 1 Map showing the location of Ejido La Victoria near El Salto
City in the State of Durango in Mexico
distinguishable in many of the sample trees. The plot
characteristics are presented in Table 1.
Competition indices
In this work the competition situation of a tree was
described with 18 different competition indices (Table 2). The first six indices (CI1–CI6) are distance-independent whereas the others (CI7 to CI18) are distancedependent ones. The indices described below were selected for evaluation in this paper due to their good
performance in previous studies (e.g. Pukkala and
Kolström 1987; Tomé and Burkhart 1989; Biging and
Dobbertin 1992, 1995; Valles et al. 1998; Bachmann
1998; Alvarez et al. 2003).
Selected distance-independent competition indices
In this study, some of the most widely used stand density
measures and indices which incorporate the vertical and
horizontal distribution of foliage were examined. CI1
represents Reineke’s (1933) stand density index, which is
based on the number of trees per hectare of plots and on
its quadratic mean diameter. In order to compute this
index the slope of the regression line (b1=1.6918) from
the stand density guide for P. cooperi developed by Cruz
and Castañeda (1999) was used. CI2 is the canopy closure that expresses the area of the crown, projected on a
horizontal plane, as a fraction of the total ground area
of the stand. CI3 represents the crown competition factor, defined as the percent of the growing area that is
occupied by the crown projection of trees assuming that
each individual is open-grown. For this study, we used
Cruz and Castañeda’s (1999) equation to predict opengrown crown width in all indices:
mcwi ¼ 0:1503di þ 2:0241
ð1Þ
1
Ejido is a land that belongs to communal groups who live and
manage their natural resources with some level of governmental
control
where mcwi is the maximum crown width (m) that could
be reached by the tree i if it was open-grown and di is its
135
Table 1 Values of the main
stand variables at the first
inventory and mean, maximum
and minimum values of the
individual tree basal area
increments for each sample plot
Site index was defined as the
dominant height at a reference
age of 50 years calculated by
using the equations proposed
by Corral et al. (2004)
Plot
1
2
3
4
5
6
7
8
Trees/ha
248
126
116
122
134
150
168
250
Basal area
(m2 ha1)
25.5
14.4
15.6
10.8
14.8
17.1
13.8
16.5
Mean dbh
(cm)
34.6
37.3
40.5
32.8
36.2
37.2
30.8
27.8
Dominant
height (m)
24.8
23.8
25.9
20.6
24.4
21.8
20.3
17.9
Site index
(m)
22.3
21.2
23.2
18.1
21.7
19.2
17.7
15.4
Individual tree
increment (cm2)
basal
area
Mean
Maximum
Minimum
115.1
111.3
87.7
123.0
104.9
96.9
77.6
61.5
292.1
230.9
209.1
267.0
201.3
191.4
179.0
168.0
6.6
34.6
21.8
37.7
26.7
21.4
8.4
8.5
Table 2 Competition indices and corresponding formulas, tested for use in the individual tree basal area growth models
Index
Source
Equation
Distance-independent competition indices:
CI1
Reineke (1933)
CI2
Canopy closure
CI3
Krajicek et al. (1961)
CI4
Wykoff et al. (1982)
CI5
Schröder and Gadow (1999)
CI6
Biging and Dobbertin (1995)
Distance-dependent competition indices:
CI7
Staebler (1951)
CI8
Gerrard (1969)
CI9
Bella (1971)
CI10
Hegyi (1974)
CI11
Alemdag (1978)
CI12
Martin and Ek (1984)
CI13
Daniels et al. (1986)
CI14
Braathe (1980, cited in Pukkala
and Kolström 1987)
CI15
Biging and Dobbertin (1992)
CI16
Biging and Dobbertin (1995)
CI17
Valles et al. (1998)
CI18
Valles et al. (1998)
10ðlog N þ1:691 log dg1:691Þ
P
( ni=1(pÆcw2i )/4)/S
P
( ni=1(pÆmcw2i)/4)/S
Pn
2
i=1(pÆdmax i )/4 = BALi
(1 [1 (BALi/G)])/RS
Pn
i=1ccj/S
P
i „ jLij
P
i „ jOij/Zi
P
i „ j(OijÆdj)/(ZiÆdi)
P
i „ jdj/(di
Æ distij)
P
i „ j{p[(distijÆdi)/(di+dj)]
P
2
P
(dj/distij)/ (dj/distij)}
i „ j(dj/di)Æ exp ((16Ædistij)/(di+dj))
P
(d2i Æn)/ i „ jd2j
P
i „ jhi/(hjÆ distij)
P
i „ jcsaj/(csai
Æ distij)
P
ccj/(cci Æ (distij+1))
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P
csaj =csai þ hj =hi =n
d
i þqh
i =distffiij þ
i6¼j
ffiffiffiffiffiffiffiffiffiffiffiffiffi
.j =d
P
nc csai
csaj p
i6¼j
i „ j
N Trees per ha in the plot, dg quadratic mean diameter (cm), n trees
per plot, cwi crown width of the subject tree (m), S plot size (m2),
mcwi maximum crown width of the subject tree estimated using
Eq. 1 (m), dmaxi dbh of the trees larger than subject tree (m), BALi
basal area of trees larger than the subject tree (m2 ha1), G basal
area of the plot (m2 ha1), RS relative spacing index of plot, Lij
distance between competitor tree and subject tree that overlap their
influence zones (m), Oij are of the influence zone overlap between
competitor and subject tree (m2), Zi area of subject tree influence
zone (m2), di subject tree dbh (cm), dj competitor tree dbh (cm),
distij distance between competitor and subject tree (m), nc number
of competitors, hi subject tree height (m), hj competitor tree height
(m), csai crown surface area of the subject tree (m2), csaj crown
surface area of the competitor (m2), exp base of the natural logarithm, ccj crown cross-sectional-area of competitor evaluated at
66% of the subject tree’s crown height from the top (m2), cci crown
cross-sectional-area of the subject tree evaluated at 66% of its
crown height from the top (m2)
diameter at breast height (cm). CI4 is the sum of the
basal areas of the trees larger than the subject tree (BAL)
proposed by Wykoff et al. (1982). C5 is a modification of
BAL, the BALMOD proposed by Schröder and Gadow
(1999), which includes the relative spacing index. CI6 is a
crown cross-sectional area index and was calculated for
each tree as the percent of the growing area that is
occupied by the crown projection of trees estimated at
66% of the subject tree’s crown length measured from
the top. Each competitor and subject tree’s contribution
to CI6 is found by taking a horizontal slice through the
canopy at this height. Competitors’ contribution to CI6
may differ with associated point of evaluation. If live
crown of a competitor tree was higher than the point of
evaluation, the contribution is then defined as the
competitor’s total live crown area, otherwise, the area at
the point of evaluation was taken into account. Afterward individual contributions are summed up and ex-
136
Fig. 2 a Crown cross-sectional area evaluated at 66% of the
subject tree’s crown height from top (CI6). The subject tree and
competitor’s contributions to CI6 are displayed with bold lines.
b Competitors are chosen with a height angle gauge of 60 from the
subject tree’s base
panded to a per hectare basis, which is then divided by
the plot area to produce an individual tree competition
index (see Fig. 2a). We selected this height because some
studies have shown that the greatest reduction in residual sum of squares of diameter and height growth
models occurs when the evaluation point is equal or
close to the one used in this study (e.g. Krumland 1982;
Biging and Dobbertin 1995; Nagel et al. 2003). As a
model for estimating the tree’s crown parameters was
not available, a paraboloid form for the tree crown of
conifer mature stands was assumed in order to compute
the crown cross-sectional surface of the trees (see Rondeux 1993, p 59).
Selected distance-dependent competition indices
Two types of distance-dependent indices were evaluated:
influence-zone overlap indices and size-ratio indices. The
first three indices CI7 (Staebler 1951), CI8 (Gerrard
1969), and CI9 (Bella 1971) are influence-zone overlap
indices which assume that the area, in a horizontal
projection, over which a tree competes for the resources
of the site, can be represented by a circle. The radius of
the circle is a function of tree size and is thought to be
equal to the expected growing space of open-grown trees
estimated using Eq. 1. Those neighboring trees that
overlap their influence zone to subject tree’s influence
zone are potential competitors.
Nine size-ratio competition indices were computed
and evaluated. These indices are based on the
hypothesis that the competition effect of a neighboring
tree increases with increasing size and decreasing distance. The first four indices use the diameter at breast
height as an indicator of size: CI10 (Hegyi 1974), CI11
(Alemdag 1978), CI12 (Martin and Ek 1984) and CI13
(Daniels et al. 1986). CI14 is a size-ratio index proposed
by Braathe (1980; cited in Pukkala and Kolström 1987)
that uses the total tree height as an indicator of size.
CI15 is a size-ratio index using the crown surface area
and it is based on the assumption that within species
growth is correlated to a tree’s crown foliage area
(Kramer 1988; Biging and Dobbertin 1992, 1995; Nagel
et al. 2002). CI16 is a size-ratio index using crown
cross-sectional area at 66% of the subject tree’s crown
length with competitors chosen with a vertical angle
gauge of 60 from the base of the subject tree (see
Fig. 2b) as was proposed by Biging and Dobbertin
(1992).
The last two size-ratio indices, C17 and C18 were
developed by Valles et al. (1998) to analyze the competition effect for P. cooperi in Mexico. From a total of
nine distance-dependent competition indices analyzed,
these two indices had the highest contribution to the
tested diameter growth model (P<0.05). The CI18 was
first proposed by Glover and Hool (1979) as a distance
independent-competition index for a basal area ratio
predictor of loblolly pine mortality. This index was
modified by Valles et al. (1998) as a distance-dependent
competition index and such modification is currently
used in the forest growth simulator SICREMAS developed for the forest region of San Dimas, Durango forest
region (Valles and Gutiérrez 2000).
Competitor selection methods
The value of a competition index for each individual tree
depends on the mathematical formulation of the relationship between the chosen variables as well as on the
method used to define neighboring trees as competitors
(Bigging and Dobbertin 1992). Various methods to
choose the trees that compete with a subject tree have
been proposed. Some of them consider as competitors
all the trees that are included within a circle with a fixed
radius whose centre is constituted by the subject tree
(Hegyi 1974). Other methods are based on variable radii,
often weighted by tree dimensions of the subject tree and
its competitors, for example dbh or height (Daniels
1976; Ford and Diggle 1981), and others combine two or
more different criteria (Bigging and Dobbertin 1992).
In this work, we tested 11 selection competitor
methods named with the codes M1 to M11. Some of
them were obtained from other studies (e.g. Pukkala
and Kolström 1987; Holmes and Reed 1991; Biging and
Dobbertin 1992; Mäkinen 1997; Alvarez et al. 2003)
and some are proposed by the authors.
The M1 identifies as competitors all those trees that
are considered in a variable plot radius sampling, the socalled ‘‘Bitterlich method’’ whose centre is the subject
tree with a basal area factor (BAF) equals to 4 m2 ha1.
Mack and Harper (1977) suggested that the competitive
influence of neighbors is greater if they are evenly distributed around a target plant, rather than aggregated
on one side. In order to take this effect into account, M2
considers as competitors all the trees selected with the
M1 method and the nearest tree that is placed inside each
one of the four quadrants defined by the cardinal points,
even though they are not within the variable plot radius
sampling. M3 selects as competitors of the subject tree its
four nearest neighboring trees.
137
Fig. 3 a For M5, competitors are chosen with a height angle of 60
from the subject tree’s crown base. b A hypothetical example
illustrating the procedure to select active competitors within the
subject tree’s influence zone, considering an angle gauge of 90.
Among the 14 potential competitors, trees numbers 1–4 are active
ones
Considering that a tree’s capacity to intercept available light has been shown to be important in growth
dynamics (Doyle 1983 in Biging and Dobbertin 1992),
the M4, first proposed by Biging and Dobbertin (1992),
is based on considering active competitors those trees
whose total height is greater than an imaginary inclined
line traced from the base of the subject tree with a 60
angle from the horizontal (Fig. 2b). The M5 is similar to
M4 but in this case the imaginary inclined line begins at
the subject tree’s live crown base. The competitors are
those trees whose height exceeds the projected angle
(Fig. 3a).
The M6 is based on the influence-zone concept
initially proposed by Staebler (1951). The potential
influence-zone of a tree is usually defined as a circle with
a radius whose value depends on the size of the subject
tree. Competition is assumed to exist when the zones of
two trees overlap. We defined as active competitors
those trees that fulfil the following condition:
distij rZi þ rZj
the first competition elimination sector and facing the reference tree is the active competitor number 2. Trees behind competitor 2 and within the second competition
elimination sector are passive competitors. The procedure
concludes when all the active competitors have been
identified. The elimination angle gauges used in this work
were 90, 75, 60, 45 and 30 for M7 to M11, respectively.
Not all the distance-dependent competition indices are
compatible with the eleven competitor selection methods.
CI7, CI8 and CI9 indices can be only computed when
combined with M6, because they are based on the influence zone concept. The other indices can be calculated
with all competitor selection methods tested in this paper.
When taking into account the different ways of
defining the competition measures, the competition situation of the trees was evaluated by 108 different variables (nine single competition measures (i.e. CI1–CI6,
CI7, CI8 and CI9) plus 99 combinations of the other
competition indices (i.e. CI10–CI18) with a competitor
selection method).
Influence of sample plots number on competition
To evaluate whether competition is effected by the
sample plot, the non-linear extra sum of squares method
(Bates and Watts 1988) was used. This method requires
the fitting of a reduced and a full model. The reduced
model corresponds to a linear model with the individual
tree basal area increment (Dgi) as the dependent variable
and the diameter at breast height (di) as independent
variable and the same set of parameters for all the
sample plots (Dgi = b0 + b1Ædi). In full model the
parameter set is different sets for each sample plot and it
is obtained by expanding each parameter (b0 and b1),
including an associated parameter and a dummy variable to differentiate the sample plot:
ð2Þ
where rZi and rZj are the influence zone radii of the
subject tree i and the competitor j respectively and distij
is the distance between the subject tree i and competitor
j. In this work we have considered as influence zone radii
the maximum value that could be reached by the tree’s
crown width if it was open-grown (Eq. 1).
The M7 to M11 take into account the competition
elimination angle concept (Lee and Gadow 1997). Each
neighbor of a given subject tree may be an active or a
passive competitor, based on a competition elimination
sector defined by a specific elimination angle. An example
illustrating the iterative procedure is presented in Fig. 3b,
using a 90 competition elimination angle. The competition
zone around the subject tree contains a defined number of
potential competitors. The nearest neighbor of the reference tree is the active competitor number 1. The trees
located behind competitor 1 and within the first competition elimination sector defined by a 90 competition
elimination angle centred in competitor number 1 are
passive competitors. The next neighbor located outside
bi ¼
8
X
bij Ij
i ¼ 1; 2
ð3Þ
j¼1
where bij are coefficients to be estimated and Ij is a
dummy variable which value is equal to 1 for sample
plot j and 0 for the remaining sample plots. The
appropriate test statistics is given by:
F ¼
(SSEðRÞ SSEðF ÞÞ=ðdfR dfF Þ
SSEðF Þ=dfF
ð4Þ
where SSE(R) is the error sum of squares of the reduced
model; SSE(F) is the error sum of squares of the full
model; dfR and dfF are the error degrees of freedom of
the reduced and full model, respectively; and F* follows
an F-distribution.
Individual tree basal area growth models
Individual tree basal area growth depends on a number
of factors such as normal section at the beginning of the
138
analyzed growth period, age, crown spread diameter,
micro-environment (i.e. site index), genetic characteristics, and the competition among trees for light, water,
nutrients, and physical growing space. Hence, an individual tree basal area growth model should consider all
these factors. However, many of the details of how these
processes are regulated are still unknown. It is difficult, if
not impossible, to generate a growth model for individual trees which includes the cited factors that affect
the tree growth. Generally researchers have accepted
that the development of individual tree growth models,
which include competition indices, is a more comprehensive attempt to estimate potential growth (Pukkala
and Kolström 1987; Biging and Dobbertin 1995; Tomé
and Burkhart 1989; Schröder and Gadow 1999; Pretzsch
2001; Nagel et al. 2002; Alvarez et al. 2003).
In the present paper, two individual tree basal area
growth models were fitted: a reduced and a complete
model. The reduced model (Eq. 5) was formulated as a
control to study the implications of modelling individual
tree basal area growth without considering competition.
This model is based on the hypothesis that individual
tree basal area growth (Dgi) is a function of the diameter
at breast height at the beginning of the analyzed growth
period (di), the site quality defined by the site index (S),
the initial stand density defined by the number of stems
per hectare (N), the initial stand basal area (G) and the
historical tree vigour (cri). If age is not available or
intentionally not used, the ratio between crown width
and tree height (crown ratio) may also be a useful predictor for describing the vigour of trees of similar stem
size (Monserud and Sterba 1996; Hökkä and Groot
1999; Schröder et al. 2002). The complete model (Eq. 6)
is similar to the reduced model, but considers the contribution of a competition index included as a new
predictor variable:
Dgi ¼ a1 þ a2 di þ a3 S þ a4 N þ a5 G þ a6 cri
R2adj ¼ 1 ðn 1Þ
n
X
n
X
ðyi ^y i Þ2 =ðn pÞ
i¼1
2
ðyi y i Þ
ð8Þ
i¼1
where yi, ^y i and y i are the measured, predicted and
average values of the dependent variable, respectively; n
is the total number of observations used to fit the model
and p is the number of model parameters.
The variance inflation factors (VIF) of all the independent variables in Eqs. 5 and 6 were calculated for
detecting multicollinearity; values up to 10 were accepted (Draper and Smith 1998; Belsey 1991; Soares and
Tomé 2001):
VIFðiÞ ¼
1
1 R2 ðiÞ
ð9Þ
where R2(i) is the multiple correlation coefficient obtained when the ith independent variable Xi is regressed
against all the remaining independent variables in the
individual tree growth model.
Finally, a partial F-test was carried out to test the
significance of including each competition index comparing the reduced model with the complete one.
Moreover, to evaluate how well a competition index
performs, the mean square error reduction (MSER)
relative to no-competition was calculated using the following expression:
MSE6
MSER ¼ 1
100
ð10Þ
MSE5
where MSE6 is the mean square error of the complete
model (Eq. 6) and MSE5 is the mean square error of the
reduced model (Eq. 5).
ð5Þ
Results and discussion
Dgi ¼ a1 þ a2 di þ a3 S þ a4 N þ a5 G
þ a6 cri þ a7 CIi
ð6Þ
where Dgi is the basal area increment of tree i (cm2),
during the analyzed period of time (5 years in this case);
di is the initial dbh of tree i (cm); S is the site index (m);
N is the initial density (stems ha1); G is the initial stand
basal area (m2 ha1); cri is the initial crown ratio of tree
i; CI is a competition index, and ai are coefficients to be
estimated.
Equations 5 and 6 were fitted by ordinary least
squares using the REG procedure of the SAS/STAT
system (SAS Institute 1999). To evaluate the performance of the fits the root mean square error (RMSE)
and the adjusted coefficient of determination (R2adj) were
estimated:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n
X
RMSE ¼
ðyi ^y i Þ2 =ðn pÞ
ð7Þ
i¼1
The results of the non-linear extra sum of squares method
(Bates and Watts 1988) revealed that the sample plot has
a significant effect on the competition (F=6.18;
P>F=0.0001). Therefore, the inclusion in Eqs. 5 and 6
of independent variables representing the initial stand
status (site index, number of stems per hectare or stand
basal area) should be considered.
The values of the VIF were always lower than 10,
except for the stand basal area (G). Therefore, this variable was finally excluded of Eq. 5 to avoid multicollinearity and to ensure that the model solution was
global rather than local.
From a total of 108 competition index varieties, only
31 increased the value of the adjusted coefficient of
determination (see Tables 3, 4). The best combinations
for each competition index are shown in Table 4.
The three first stand level density measures (CI1–CI3)
did not show a significant reduction of MSE when
compared with Eq. 5. These results come across with
139
Table 3 Mean square error
reduction of the combinations
of distance-dependent
competition indices and
competitor selection methods
that showed an improvement of
the adjusted coefficient of
correlation comparing to using
no competition index (Eq. 5)
The line indicates no improvement
Table 4 Contribution of
competition indices to
individual tree basal area
growth models (multiple linear
regressions)
From the 108 fitted models,
only the best variation of each
competition index is shown here
a
Significant at 0.05 level;
b
Significant at 0.01 level; R2E5
and R2E6 are the adjusted coefficient of determination of the
reduced (Eq. 5) and the
complete model (Eq. 6),
respectively
Distance
Competitor selection method
dependent
M2
M3
M4 M5
M1
M6
M7
M8
M9
M10
M11
CI10
CI11
CI12
CI13
CI14
CI15
CI16
CI17
CI18
3.1357
0.1764
0.0786
–
–
–
–
–
–
–
–
–
–
–
–
0.5613
–
–
–
–
–
–
–
–
1.7280
–
–
–
0.0322
–
–
–
–
1.9680
–
–
–
–
–
–
–
–
1.9735
0.4698
0.0369
–
–
–
–
0.0315
0.0421
2.1844
0.5512
–
1.3674
–
–
–
–
–
–
–
–
1.1662
–
0.4025
–
–
–
0.7074
–
–
–
–
–
–
–
–
1.7285
–
–
–
–
–
–
–
–
–
–
–
1.1723
–
0.4132
–
–
–
0.7624
–
Competition
index
n
R2E5
R2E6
Partial-F
(index term)
RMSE
MSER
(%)
CI1
CI2
CI3
CI4
CI5
CI6
CI7M6
CI8M6
CI9M6
CI10M11
CI11M6
CI12M6
CI13M5
CI15M11
CI16M11
CI17M11
CI18M11
310
310
310
310
310
310
289
289
289
310
289
289
310
310
310
310
310
0.6792
0.6792
0.6792
0.6792
0.6792
0.6792
0.6823
0.6823
0.6823
0.6792
0.6823
0.6823
0.6792
0.6792
0.6792
0.6792
0.6792
0.6795
0.6804
0.6805
0.6843
0.6868
0.6816
0.6842
0.6861
0.6838
0.6801
0.6934
0.6840
0.6816
0.6802
0.6803
0.6871
0.6821
0.26
1.12
1.25
4.94a
7.37b
2.32
1.73
3.43
1.37
1.01
10.21b
1.52
2.29
0.92
1.03
7.65b
2.73
8.08
8.06
8.05
8.01
7.98
8.04
7.92
7.90
7.93
8.06
7.81
7.93
8.05
8.06
8.06
7.98
8.04
0.0204
0.0786
0.0927
1.2734
2.0573
0.4221
0.2444
0.8317
0.1162
0.0369
3.1357
0.1764
0.4132
0.0315
0.0421
2.1844
0.5512
those found by Biging and Dobbertin (1995) who in
general, did not find any improvement for the diameter
squared and height growth models, they tested, when
traditional stand density measures were used as competition indices. Thus, the results found in this study
make evident the main problem of the use of the stand
variables as competition indices, i.e. all counted trees
contribute equally to the competition estimate, despite
size or proximity (Opie 1968; Moore et al. 1973; Alemdag 1978; Mäkinen 1997).
On the other hand, the distance-independent indices
that consider tree basal area or crown parameters of selected competitors around a subject tree (CI4–CI6)
showed a higher reduction of the MSEs comparing to
using no competition index, specially those indices based
on the sum of the basal areas of the trees larger than the
subject tree (BAL and BALMOD). The contributions of
the distance-independent indices: BAL (CI4) and BALMOD (CI5) showed highly significant values in the partial F-test (Table 4). These results are consistent with
those obtained in similar studies with other conifer species (e.g. Biging and Dobbertin 1985; Álvarez et al. 2003).
As can been seen in Table 4, none of the selected
influence-zone overlap competition indices (CI7, CI8 and
CI9) showed a significant contribution (Partial-F test) to
the growth model compared to using no competition
index.
The results of the size-ratio indices that use the dbh as
indicator of size (CI10, CI11, CI12 and CI13) depend on
the use of the competitor selection method. The best
results were obtained for the index proposed by Alemdag (CI11) combined with the M6 that is based on the
influence-zone concept defined as a circle with a radius
equal to the crown width of an open grown tree. This
competitor selection method showed also the best results
for indices CI12 and CI13, although, in these cases the
results of partial-F test were not significant.
The size-ratio index based on the total height (CI14)
did not show any contribution to the individual tree
basal area growth model, no matter which competitor
selection method was used. These results differ with
those obtained with size-ratio indices based on dbh
dimensions when the partial contribution of competition
index depends on the competitor selection method. In
order to find an explanation for this effect, the spatial
quantities diameter and height differentiation were calculated, using the variable size differentiation proposed
by Gadow and Füldner (1995) 2 based on the three
nearest neighbors of a given reference tree. Average
2
The differentiation of a size variable (Ti) for a given tree i and its n
nearest
P neighbors j (j=1, ..., n) is defined as follows: Ti = 1 (1/
n) nj=1 (min (Xi, Xj))/(max (Xi, Xj)) with X = a size variable and
0 £ Ti £ 1
140
values of size differentiation of 0.2413 and 0.1342 for
dbh and total height were observed respectively, indicating a greater homogeneity in total height when
compared with dbh. Hence, the standard deviation of
the size-ratio indices based on total height for any
competitor selection method is smaller than those obtained for size-ratio indices based on dbh.
Biging and Dobbertin’s size-ratio indices that use the
crown dimensions as indicator of size (CI15 and CI16)
did not present an improvement of the adjusted coefficient of determination of Eq. 5, except when they are
combined with the M11 competitor selection method
which corresponds to the competition elimination angle
concept, using an angle gauge of 30. However, the results of the partial F-test indicate that these competition
indices varieties do not contribute significantly to the
individual tree basal area growth when they are included
as competition indices.
The distance-dependent competition index CI17
proposed by Valles et al. (1998) for P. cooperi showed,
in general, good results among the competition indices
analyzed. The inclusion of this competition index in
Eq. 5 results in reduction of the MSE for 8 of the 11
competitor selection methods tested in this study. The
best combination for this index in terms of Partial-F
test (7.65; significant at a 99% level) was obtained
using competition elimination angle of 30 (M11). On
the other hand, the best combination for competition
index CI18 was found to be CI18M11. This competition
index was proposed also by Valles et al. (1998) and is
currently used in the forest growth simulator SICREMAS developed for the forest region of San Dimas,
Durango. This variety corresponds as well to the M11
competitor selection method which derived from the
competition elimination angle concept, using an angle
gauge of 30. However, CI18M11 variety performed
slightly worse than the best distance-independent indices (CI4 and CI5) or some of the size-ratio indices
based on dbh (CI11) or based on dbh and total height
(CI17).
In relation to the competitor selection methods, in
general, the best results were obtained with those based
on the influence-zone concept defined as a circle with a
radius equal to the crown width of an open grown tree
(M6) or based on the competition elimination angle concept, using an angle gauge of 30 (Table 3). Therefore,
the use of small angle gauges for mature stands of P.
cooperii with low density is more suitable to select an
adequate number of competitor trees since more competitor trees are chosen than using large angle gauges.
Alvarez et al. (2003) found also good results for the
competition elimination angle concept but using angle
gauges of 45 or 60 in their competition study with
young Pinus radiata stands.
In general, the selected distance-dependent competition indices performed best, in terms of adjusted R2,
RMSE, partial F-tests and the mean square reduction
(MSER) relative to no competition when they are
combined with the influence-zone concept (M6) or
competition elimination angle concept (M11; angle of
30) competitor selection methods.
Seven of the selected distance-dependent indices
(CI10, CI12, CI13, CI14, CI15, CI16 and CI18) did not show
a significant contribution to the individual tree basal
area growth model (see Table 4). However, taking into
account that the sample plot data are derived from
mature and thinned stands (with less than
238 stems ha1) where the competition effect is not
great, further studies in younger stands are needed in
order to evaluate their suitable contribution to individual basal area growth models for P. cooperi.
As mentioned above, the distance-dependent index
proposed by Alemdag (1978) combined with a competitor selection method based on the influence-zone
concept (CI11M6) and the distance-dependent index
proposed by Valles et al. (1998) combined with a
competition elimination angle of 30 (CI17M11) performed slightly better than BAL (CI4) and BALMOD
(CI5) indices. However, the BAL and BALMOD indices can be derived using variables that are readily
available (number of trees per hectare, basal area and
dominant height) for each sub-compartment in the
study area. Since one of the objectives of this work was
to find a competition index which had a significant
contribution to the individual tree basal area growth
model and does not require too many field measurements, the BALMOD is recommended to be incorporated into an individual tree basal area growth model
for P. cooperi in the forest region of El Salto, Durango
Mexico. Figure 4 shows a scatterplot illustrating the
relationship between individual tree basal area growth
and BALMOD-index.
The BALMOD index showed important reductions
in the MSEs for growth models in previous competition
studies of Pinus pinaster (5.13%; Schröder and Gadow
1999) and P. radiata (24.3%; Alvarez et al. 2003) in
Spain. In these studies, the evaluated stands were
younger and with a higher number of stems per hectare
than the ones used in our analysis. Hence, a considerable
improvement of the individual tree basal area growth
model for P. cooperi could be attained using the
Fig. 4 The relationship between individual tree basal area growth
and BALMOD-index
141
BALMOD index with a more complete database
including more combinations of different ages, site
qualities and stand densities.
Conclusions
The distance-independent competition indices BAL
(CI4) and BALMOD (CI5) performed as well as the best
distance-dependent competition indices (CI11M6, and
CI17M11) based on the mean square error reduction they
induced when added to an individual basal area growth
model without competition effect. When these four
competition indices were included in the model, the increase in the adjusted multiple coefficient of determination was not very high in magnitude. However, the
variables BALMOD (CI5), CI11M6, and CI17M11 were
highly significant while BAL (CI4) was significant. The
results indicate that, for P. cooperi growing in even-aged
stands, the BALMOD index is the most suitable one to
be incorporated into an individual tree basal area
growth model for the forest region of El Salto, Durango,
Mexico. This competition index performs as well as the
other indices analyzed, including index (CI18) proposed
by Valles et al. (1998), for this pine species in the forest
region of San Dimas, Durango. The BALMOD index is
much easier to calculate using data that are usually
available.
When comparing the different competitor-selection
methods, we conclude that the influence-zone concept
(M6) defined by a circle with a radius equal to the
crown width of an open grown tree and the competition elimination angle concept using an angle gauge of
30 are the most reasonable criteria for defining
neighboring trees as active competitors for this species,
based on the data used. However, the competition
elimination angle concept has the advantage that is very
easy to use in the field and for other stand conditions,
different angles could be tested to take into account
the stand density and thus to control the number of
competitors.
In the pine stands evaluated in this study, about 33%
of the individual tree basal area growth variation could
not be explained. Part of this variation may be caused by
genetic differences between trees (Perry 1985) or by the
spatial variation in the site attributes (Pukkala and
Kolström 1987) that are difficult to model. An unexplained part may be due to competition effects that are
not expressed by the chosen indices. Therefore, as has
been mentioned, additional research on tree competition
effects is required for a better understanding of the
competition dynamics and to search for a more physiological competition measure.
Acknowledgements Support for this research was provided through
a cooperative agreement between the Mexican National Council
for Science and Technology (project: 41181-Z) and the German
Academic Exchange Service. We thank Prof. von Gadow for
revising the English. Two anonymous reviewers also provided
useful comments.
References
Alemdag PG (1978) Evaluation of some competition indexes for
the prediction of diameter increment in planted white spruce. F
Man Ins Inf Rep Fmr-x-108, p 39
Alvarez MF, Barrio M, Gorgoso F, Alvarez JG (2003) Influencia
de la competencia en el crecimiento en sección en Pinus radiata
D. Don Invest Agrar Sist Rec For 12(2):25–35
Bachmann M (1998) Indizes zur Erfassung der Konkurrenz von
Einzelbäumen.
Methodische
Untersuchung
in
Bergmischwäldern. Forstliche Forschungsberichte München, Nr.
171, 245 S
Bates DM, Watts DG (1988) Nonlinear regression analysis and its
applications. Wiley, New York
Bella IE (1971) A new competition model for individual trees. For
Sci 17:364–372
Belsey DA (1991) Conditioning diagnostic, collinearity and weak
data in regression. Wiley, New York
Biging GS, Dobbertin M (1992) A comparison of distance-dependent competition measures for height and basal area growth of
individual conifer trees. For Sci 38(3):695–720
Biging GS, Dobbertin M (1995) Evaluation of competition indices
in individual tree growth models. For Sci 41(2):360–377
Brand DG, Magnussen S (1988) Asymmetric, two-sided competition in even-aged monocultures of red pine. Can J For Res
18:901–910
Clark PJ, Evans FC (1954) Distance to nearest neighbor as a
measure of spatial relationships in populations. Ecology
35:445–453
Corral RJ, Álvarez González JG, Ruı́z González A, Gadow Kv
(2004) Compatible height and site index models for five pine
species in El Salto, Durango (Mexico). For Ecol Manag
201:145–160
Cruz CF, Castañeda AJ (1999) Guia de densidad para Pinus cooperi var. Blancoi del ‘‘Ejido La Victoria’’ Pueblo Nuevo Durango. X Congreso de la Dirección General de Educación
Tecnológica Agropecuaria. Oaxaca, Mexico
Daniels RF (1976) Simple competition indices and their correlation
with annual lobolly pine tree growth. For Sci 22:454–456
Daniels RF, Burkhart HE, Clason TR (1986) A comparison of
competition measures for predicting growth of loblolly pine
trees. Can J For Res 16:1230–1237
Draper NR, Smith H (1998) Applied regression analysis, 3rd edn.
Wiley, New York
Ford ED, Diggle PJ (1981) Competition for light in a plant
monoculture modelled as a spatial stochastic process. Ann Bot
48:481–500
Gadow Kv, Füldner K (1995) Zur Beschreibung forstlicher Eingriffe. Forstw Cbl 114:151–159
Gadow Kv, Hui GY (1999) Modelling forest development. Kluwer
Academic Publishers, Dordrecht
Garcia AA, Gonzalez EMS (1998) Pináceas de Durango. Instituto
de Ecologı́a-CIIDIR-IPN. Durango, Dgo. México
Gerrard DI (1969) Competition quotient: a new measure for the
competition affecting individual forest trees. Mich State Univ
Agric Res Station Res Bull 20:1–32
Glover GR, Hool JN (1979) A basal area predictor of loblolly pine
plantation mortality. For Sci 25:275–282
Hegyi F (1974) A simulation model for managing jackpine stands.
In: Fries J (ed) Growth models for tree and stand simulation.
Proceedings of IUFRO meeting S4.01.04, Royal College of
Forestry, Stockolm, pp 74–90
Hökkä H, Groot A (1999) An individual-tree basal area growth
model for black spruce in second-growth peatland stands. Can J
For Res 29:621–626
Holmes MJ, Reed DD (1991) Competition indices for mixed species Northern Hardwoods. For Sci 37(5):1.338–1.349
Krajiceck JE, Brink-Man KA, Gingrich F (1961) Crown competition a measure of density. For Sci 7:35–42
Kramer H (1988) Waldwachstumslehre. Verlag Paul Parey, Hamburg
142
Krumland B (1982) A tree-based forest yield projection system for
the north coast region of California. PhD Thesis, University Of
California, Berkeley
Lee DT, Gadow Kv (1997) Iterative bestimmung der konkurrenzbäume in Pinus densiflora Beständen. AFJZ 168(3/
4):41–44
Lorimer CG (1983) Test of age-independent competition indices for
individual trees in natural hardwood stands. For Ecol Manag
6:343–360
Mack RN, Harper JL (1977) Interference in dune annuals: spatial
pattern and neighborhood effects. J Ecol 65:345–364
Mäkinen H (1997) Possibilities of competition indices to describe
competitive differences between Scots pine families. Silva Fenn
31:43–52
Martin GL, Ek AR (1984) A comparison of competition measures
and growth models for predicting plantation red pine diameter
and height growth. For Sci 30(3):731–743
Monserud RA, Sterba H (1996) A basal area increment model for
individual trees growing in even and uneven-aged forest stands
in Austria. For Ecol Manag 80:57–80
Moore JA, Budelsky CA, Schlesinger RC (1973) A new index
representing individual tree competitive status. Can J For Res
3:495–500
Nagel J, Albert M, Schmidt M (2002) Das waldbauliche Prognoseund Entscheidungsmodell BWINPro 6.1. Forst u Holz 57:486–
493
Nagel J, Albert M, Schmidt M (2003) BWINpro: program zur
Bestandesanalyse und Prognose. Handbuch zur Version 6.1,
Göttingen
Opie JE (1968) Predictability of individual tree growth using various definitions of competing basal area. For Sci 14:314–323
Perry DA (1985) The competition process in forest stands. In:
Cannell MGR, Jackson JE (eds) Trees as crop plants. Institute
of Terrestrial Ecology. Natural Environment Research Council,
pp 481–506
Pretzsch H (2001) Modellierung des Waldwachstums. Parey-Verlag, Berlin
Pretzsch H (2002) Grundlage der Waldwachstumforschung. PareyVerlag, Berlin
Pretzsch H, Biber P, Dursky J (2002) The single tree-based stand
simulator SILVA: construction, application and evaluation.
For Ecol Manag 162:3–21
Pukkala T, Kolström T (1987) Competition indices and the prediction of radial growth in Scots pine. Silva Fenn 21:55–67
Reineke LH (1933) Perfecting a stand density index for even-aged
forest. J For Res 46:627–638
Rodrı́guez FC (1987) Development of a competition index for Pinus montezumae Lamb. in a temperate Forest of Mexico. PhD
Thesis. Yale University School of For., p 200
Romero GYE (1993) Análisis del crecimiento de Pinus patula Schl.
et Cham. en diferentes niveles de competencia intraespecı́fica,
en Huayacocotla, Ver. Tesis M.C. División de Ciencias Forestales. Chapingo, Mexico
Rondeux J (1993) La mesure des arbres et des peuplements forestiers. Les Presses Agronomiques de Gembloux
SAS Institute Inc (1999) SAS/STATTM User’s Guide, Relase 8.0
Edition. Cary, USA
Schröder J, Gadow Kv (1999) Testing a new competition index for
Maritime pine in northwestern Spain. Can J For Res 29:280–283
Schröder J, Rodriguez R, Vega G (2002) An age-independent basal
area increment model for Maritime pine trees in northwestern
Spain. For Ecol Manag 157:55–64
Soares P, Tomé M (2001) A tree crown ratio prediction equation
for eucalypt plantations. Ann For Sci 58:193–202
Staebler GR (1951) Growth an spacing in an even-aged stand of
Douglas-fir. Master’s thesis, University of Michigan
Tome M, Burkhart HE (1989) Distance-dependent competition
measures for predicting growth of individual trees. For Sci
35(3):816–831
Torres JM (2000) El área de ocupación promedio: Una medida de
competencia independiente de la distancia. Agrociencia
34(4):485–495
Valles G, Gutiérrez IF (2000) Sistema de ecuaciones del simulador
de crecimiento maderable para la región de San Dimas, Durango. Centro de Investigación Regional Norte Centro. Publicación especial Num. 14. Durango, México
Valles GAG, Torres RJM, Velázquez MA, Rodrı́guez FC (1998)
Relación de nueve ı́ndices de competencia con el crecimiento en
diámetro de Pinus cooperi Blanco. Agrociencia 32(3):255–260
Van Laar A, Akca A (1997) Forest mensuration. Cuvillier, Göttingen, 418 p
Wykoff WR, Crookston NL, Stage AR (1982) User’s guide to the
stand prognosis model. Intermountain Forest And Range
Experiment Station, General Technical Report INT-133