Stand canopy cover is one of the most important structural characteristics
of a forest stand. The state of shading of stand soil by the stand canopy is
called stand canopy cover (Saatcioglu, 1976). Canopy cover
is determined with the percent of a forest area occupied by the vertical projections
of tree crowns (Avery and Burkhart, 1994). Canopy cover
is strongly related to photosynthetically active radiation and overstory leaf
area index (Buckley et al., 1999). Stand canopy
cover has considerable effects on soil, stem and crown development and natural
regeneration (Genc, 2004). In this respect, it is quite
important to manage stand canopy cover in a way suitable for the objective in
stand tending and natural regeneration activities. On the other hand, canopy
cover is a characteristic of which determination is required for describing
stand structure. In addition, canopy cover can be used as a variable to predict
stand volume from aerial photographs (Eraslan and Sad, 1993;
Percent canopy cover can be estimated by ground-based measurements, aerial
photographs or satellite images. Determining canopy cover by ground-based measurements
is relatively difficult and time-consuming. Therefore, ground-based canopy cover
measurements may be preferred when accuracy is an important objective (Fiala
et al., 2006). Different methods such as ocular estimation (Genc,
2004; Korhonen et al., 2006), Cajanus tube,
LAI-2000 plant canopy analyzer instrument (Rautiainen et
al., 2005), line-intercept, spherical densiometer, moosehorn and hemispherical
photography (Fiala et al., 2006) could be used
for determining canopy cover as ground-based. In this respect, by comparing
with each other the results obtained by different methods, the selection of
the method more suitable to the objective has importance.
One of the methods that can be used in ground based estimations of stand canopy
cover is also estimation method by crown projection map with scale (Genc,
2004). In the studies made to determine stand structures, stand profiles
and crown projections for the
sample plots selected from the stands have been drawn (Bozkus,
1987; Ozalp, 1989; Demirci, 1991)
and percent canopy covers of these stands have generally been estimated by eye
in the office according to the crown projections drawn for the sample plots.
However, because graphic methods have also developed with the development of
computer programs today, it is also necessary to investigate the usability of
graphic methods besides the ocular estimation method in determination of percent
canopy covers on the crown projection diagrams with scale.
In this study, the possibilities of estimation of percent canopy cover by different methods on the crown projection diagrams with scale were investigated in the sample of Lebanon cedar (Cedrus libani A. Rich.) stands of the Yavsan Mountain, Kahramanmaras. For this, ocular estimation values of different persons and estimation values of different graphic methods were compared with each other; thus, it was aimed to compare the different methods with each other.
MATERIALS AND METHODS
Six sample-plots of 10x50 m each were selected in the direction perpendicular to the contour lines from natural pure and mixed stands of Lebanon cedar at generally pole-thin tree stages in the Yavsan Mountain (1964 m), Kahramanmaras, southern Turkey (Table 1, 2). Each sample plot was divided into five parcels of 10x10 m (100 m2), the locations of trees, of which heights were ≥5 m, in each parcel were marked on the graph paper as with coordinate and crown widths were measured from four different directions. Crown projections of the trees were drawn on the graph paper with a scale of 1:200 in the office and then were drawn on the tracing paper from this paper. After photocopying the tracing papers, the photocopies were transferred into computer through a scanner, and percent canopy cover of each parcel was tried to be determined from over 1.0 on the obtained digital images (jpeg).
It was benefited from ocular estimation method and three different graphic methods in determining percent canopy covers of the parcels. These methods are explained below:
Ocular estimation method: Percent canopy covers in the parcels were estimated by two experienced persons (A and B) and one less experienced person (C) with a precision of 5% by examining crown projection areas and empty areas in the parcels directly on the photocopy papers (A4).
||Some site characteristics of the sample plots
|*S: Lebanon cedar (Cedrus libani), G: Cilician fir
(Abies cilicica ssp. cilicica), Ck: Crimean pine (Pinus
nigra ssp. pallasiana)
||Some stand characteristics in the sample plots
Digital square grid method: Crown projection areas in the parcels were filled with black colour through PhotoFiltre (ver. 6.3.2) program on the digital images, then digital square grids were generated on the parcels, and squares belonging to the empty areas in these grids were counted by us. During the counts, half and over empty squares were considered. The number of the total squares of the empty areas was proportioned to the number of the total squares of that parcel, the obtained value was subtracted from 1.0, thus percent canopy cover of that parcel was found.
Polygon method: Polygons were drawn by enclosing the surroundings of empty areas in the parcels with ImageTool (ver. 3.00) program on the digital images and the areas of these polygons were calculated in pixels2 by the program. The total pixels2 value of empty areas in the parcel was proportioned to the total pixels2 value of that parcel, and the obtained value was subtracted from 1.0, thus percent canopy cover of each parcel was determined.
Binary image method: Crown projection areas in the parcels were filled with black colour through PhotoFiltre program on the digital images, then these images were converted to binary image (tif) by using ImageJ (ver. 1.41o) program; in general, crown projection areas were covered with white colour while empty areas were covered with black colour in the binary image. When crown projection areas were less than empty areas, a complete opposite of this coloration occurred. The numbers of black and white pixels in the binary images and their proportions to the total number of pixels in the parcel were calculated by ImageTool program. Thus, the proportion of generally white or sometimes black pixels gave percent canopy cover in the parcels.
After percent canopy covers of five parcels in each sample plot were estimated by the different methods, the mean of five values obtained by each method was calculated and the mean percent canopy covers in the sample plots were found separately according to the methods.
One-way analysis of variance (ANOVA) (Kalipsiz, 1981)
was performed to determine whether there was a statistically significant difference
in terms of the mean percent canopy cover among the methods for each sample
plot, the data were arcsin(p)1/2 transformed before the analyses.
In addition, correlation analysis (Kalipsiz, 1981) was
applied to determine whether there was a statistically significant relationship
in terms of percent canopy cover among the methods. For this, transformed percent
canopy cover values of a total of N=30 parcels in six sample plots were used
and Pearson correlation coefficients (R) were calculated. In all statistical
analyses, a confidence level of p = 0.05 was used for statistical significance
and the analyses were carried out by using SPSS 16.0 package (SPSS
The mean percent canopy covers of the sample plots according to the methods
varied between 0.61-0.71 for sample plot I, 0.48-0.54 for sample plot II, 0.66-0.72
for sample plot III, 0.72-0.85 for sample plot IV, 0.59-0.65 for sample plot
V and 0.59-0.65 for sample plot VI (Table 3). Thus, there
were differences ranging from 6 to 13% according to the methods in the mean
percent canopy covers of the sample plots. The highest difference was between
the binary image method and Cs ocular estimation with 13% in sample plot
The general mean percent canopy cover of all the sample plots varied also between 0.61 (the binary image method) and 0.66 (As and Cs ocular estimation) according to the methods (Table 3). Thus, the highest difference among the general mean percent canopy covers of all the sample plots was between the methods mentioned above with 5%.
According to the general mean values in Table 3, in the ocular estimation method, it is seen that there was a difference up to 4.0% among As, Bs and Cs ocular estimation values, As and Cs ocular estimation values were the same with each other, and Bs ocular estimation values were relatively lower. There was also a difference up to 4% among the values of the graphic methods, the methods of polygon and digital square grid gave results close to each other, and the binary image method yielded relatively lower values. Furthermore, when comparing the mean of the ocular estimation method (0.65) and the mean of the graphic methods (0.63), 2% higher values were obtained with the ocular estimation method in the general mean.
The results of one-way ANOVA showed that there was no statistically significant difference (p > 0.05) in terms of the mean percent canopy cover among the methods for all the sample plots [sample plot I (F = 0.968; p = 0.457), sample plot II (F = 0.710; p = 0.622), sample plot III (F = 0.530; p = 0.752), sample plot IV (F = 0.879; p = 0.510), sample plot V (F = 0.155; p = 0.977) and sample plot VI (F = 0.043; p = 0.999)]. Thus, it is understood that there was also no statistically significant difference in terms of percent canopy cover both among As, Bs and Cs ocular estimations and among the graphic methods.
The results of correlation analysis revealed that there were very strong positive
relationships (R>0.90; p<0.01) in terms of percent canopy cover among
all the methods (Table 4). Pearson correlation coefficients
(R) varied between 0.903 and 0.944 in the relationships among As,Bs
and Cs ocular estimation; and also varied between 0.993 and 0.996 in the
relationships among the methods of digital square grid, polygon and binary image.
|| The mean percent canopy covers of the sample plots according
to the methods (Mean±standard deviation)
||The results of correlation analysis of relationships among
Thus, it is seen that the relationships among the graphic methods were very
stronger and the strongest relationship was between the methods of digital square
grid and polygon (R = 0.996).
In the estimation of percent canopy covers in the parcels, the simplest and relatively the fastest method was the ocular estimation method, followed by the methods of binary image, digital square grid and polygon, respectively. While the digital square grid and polygon methods were carried out by using the only one computer program, the binary image method was performed with three different programs by us.
While the highest difference determined according to the methods in the mean
percent canopy covers of the sample plots was 13% (sample plot IV), the highest
difference among the general mean percent canopy covers was found to be 5%.
Namely, the differences among percent canopy covers obtained with the methods
decreased in the general mean. But, because canopy cover of each sample plot
or stand has separately importance, this maximum difference of 13% occurred
according to the methods should be taken into account, especially in precision
studies. Korhonen et al. (2006) also stated that
different techniques yielded considerably different canopy cover estimations.
That there was no statistically significant difference (p>0.05) in terms of the mean percent canopy cover in the sample plots among the methods indicates that percent canopy cover could be determined on the crown projection diagrams with scale by any of the methods mentioned. In this way, the selection of the method more suitable to the objective from the methods explained gains importance here.
There was no statistically significant difference (p>0.05) in terms of percent
canopy cover between ocular estimations of the experienced persons (A and B)
and less experienced person (C). This situation may result from making the estimations
in the office and on the drawn crown projections. Therefore, the obtaining of
different results according to the persons could be expected in ocular estimations
to be made in the field. Indeed, Korhonen et al.
(2006) determined that the ocular estimations achieved in the field without
proper training might be seriously biased in either direction.
In the study, although the ocular estimation method gave relatively fast results,
the estimations in this method were able to be made initially with at least
an error of 5%. Similarly, Avsar (1999), Tonguc
(2003) and Ayyildiz (2003) estimated stand canopy
covers on the crown projections with scale by using the ocular estimation method
and with a precision of 10%. In this respect, the error rate in the ocular estimation
method could be more than the graphic methods. Moreover, in the present study,
2% higher values were obtained with the ocular estimation method than the graphic
methods in the general mean. Korhonen et al. (2006)
also reported that, in general, labour intensive techniques (the Cajanus tube,
line intersect sampling) provide unbiased and more precise estimations, while
the estimations provided by fast techniques (digital photographs, ocular estimation)
have larger variances and may also be seriously biased.
There was no statistically significant difference (p>0.05) in terms of percent
canopy cover among the graphic methods as well. However, the polygon method
(0.64) and the digital square grid method (0.65) gave results closer to each
other than the binary image method (0.61) in the general mean. Avery
and Burkhart (1994) also noted that the dot grid estimation technique has
the virtue of producing a reasonable degree of consistency among various photo
interpreters. On the other hand, in the study, the binary image method was the
method in which the general mean percent canopy cover was the lowest. It was
determined that this situation resulted from the different perception of the
signs showing the center of tree and the pixels in sidelines of the crown projection
area in the drawn crown projection diagrams. In order to overcome this, making
some additional correction operations is required on the existing digital images.
There were very strong positive relationships (R>0.90; p<0.01) in terms
of percent canopy cover among all the methods. But, the relationships among
the graphic methods were very stronger, and the strongest relationship was between
the methods of digital square grid and polygon (R = 0.996). This revealed that
percent canopy covers obtained by the graphic methods had a very higher linear
relationship with each other than those obtained by the ocular estimation method.
Rautiainen et al. (2005) also determined that
the Cajanus tube yielded slightly larger canopy cover values than the LAI-2000
plant canopy analyzer instrument in the field, but the values were nevertheless
in good agreement.
In spite of at least an error of 5% at first, the ocular estimation method
is a method which is relatively simple and gives fast results. In this respect,
if the time is limited, the ocular estimation method can be preferred in determining
percent canopy covers on the crown projections with scale. If the time is not
limited and more precise results are needed, the methods of digital square grid
or polygon from the graphic methods can be suggested. Despite seeming like a
complex method and giving lower percent canopy covers, it can be said that the
binary image method is an alternative method which is quite fast and to be achieve
good results in case of making some additional corrections on the digital images,