Open Access is an initiative that aims to make scientific research freely available to all. To date our community has made over 100 million downloads. It’s based on principles of collaboration, unobstructed discovery, and, most importantly, scientific progression. As PhD students, we found it difficult to access the research we needed, so we decided to create a new Open Access publisher that levels the playing field for scientists across the world. How? By making research easy to access, and puts the academic needs of the researchers before the business interests of publishers.
We are a community of more than 103,000 authors and editors from 3,291 institutions spanning 160 countries, including Nobel Prize winners and some of the world’s most-cited researchers. Publishing on IntechOpen allows authors to earn citations and find new collaborators, meaning more people see your work not only from your own field of study, but from other related fields too.
To purchase hard copies of this book, please contact the representative in India:
CBS Publishers & Distributors Pvt. Ltd.
www.cbspd.com
|
customercare@cbspd.com
We introduce a procedure to predict the vegetation shift using traditional phytosociological survey (cover data). The cover value is generally obtained for each layer of the layered plant community, but usually maximum cover value over the layers used for the vegetation classification and recognition (C-max procedure). As an ameliorate procedure, we propose the procedure of every coverage of all layers used to evaluate vegetation shift (C-all procedure). The C-all procedure enables us to embrace the information on vertical gradient of species distribution in the surveyed communities. In the case of our observations and analyses, tree species with smaller (or no) cover in the upper layer but greater cover in the lower layer can be dominant in the upper layer in the future, resulting in vegetation shift (changes in dominant species of the community). Every general community analysis (cluster analysis, INSPAN, and TWINSPAN) followed by C-all procedure supports such prediction for some types of Japanese forests. In the forests, changes in species composition have been conventionally predicted by measuring the trunk diameter and height of trees. Our proposal suggests that traditional phytosociological survey is also convenient for studying forest succession and regeneration.
Faculty of Design Technology, Osaka Sangyo University, Osaka, Japan
Yui Oyake
Faculty of Engineering and Design, Kagawa University, Kagawa, Japan
Tsuyoshi Kobayashi
Faculty of Agriculture, Kagawa University, Kagawa, Japan
*Address all correspondence to: shimano@est.osaka-sandai.ac.jp
1. Introduction
In this article, we would like to describe our method for predicting vegetation changes using phytosociological vegetation data, which is usually collected for vegetation classification. Our hope is that the accumulated data of phytosociological vegetation surveys will provide a method for not only community classification but also a method to know vegetation dynamics and future prediction of vegetation in a certain procedure. For this purpose, we introduce ameliorated analyzing procedure using the information on coverage of plants collected from the layered forests at each layer (hierarchical data) and ways to estimate vegetation shifts such as forest regeneration and succession.
We explain how phytosociological vegetation data are collected in short. Within a certain area selected as a study site for vegetation survey where the occurrence of plants in the community is expected to be uniform, which often means a sample “plot” (quadrat) extracted from a homogeneous stand in an area of a site, we record the existing species in the plot and the above-ground dominance in each layer (along a vertical direction of the community; e.g., tree layer, sub-tree layer, shrub layer, herb layer) over the plot. Then, the abundance of each species is measured, and the dominant species are evaluated.
2.1 Measuring the coverage
Various indices of dominance can be considered, such as the number of individuals, plant volumes, cover, etc. As a traditional standard, the “rank of cover rate” proposed by Braun-Blanquet [1], indicates how much of the projected area of target vegetation is occupied by each species composing the community. The coverage ranks as abundance of each species are 5, 4, 3, 2, 1, and +:
Here, even if the above-ground parts of a plant (crown and/or trunk of a tree) traverse over the tree layer and the sub-tall-tree layer or lower layers, the cover value of that plant is aggregated as a value of the upper layer and not as other layers. For the plants of other layers, the cover of the plants constituting each layer is determined based on the vertical distribution of plants. This method has been widely accepted and a vast amount of data has been collected around the world.
It is important to note that in the layered forests in the temperate regions, for example, researchers record data according to the following hierarchical levels: the tall-tree layer, which constitutes the forest canopy (overstory); the sub-tall-tree layer, which is slightly lower in height than the tall-tree layer; the shrub layer, which is about 4–5 m high; and the shrub layer, which contains herbaceous plants (including grasses, sedges, sometimes ferns, lianas and so on), seedlings, and juveniles of trees and shrubs less than 1 m in height. This work means that the hierarchical structure of the community is faithfully recorded. Not only in the different types of forests in other regions but also in the herbaceous communities and grasslands are such hierarchical data collections useful.
2.2 General uses of coverage (C-max procedure) and demerits
The classical classification and recognition of community type are based on the concept of “Environmental Diagnosis”. A recognized vegetation provides an indicator of the climate and other regional environmental traits, such as dryness or wetness of the soil, which has been established in the plant community.
Generally, in such Environmental Diagnostic studies, the data taken from the idea of Braun-Blanquet are carefully organized and analyzed using the method of Mueller-Dombois and Ellenberg [2], which is the central text in the field. In their methods, data recorded by tall-tree, sub-tall-tree, shrub, and herbaceous strata, in other words, along a developmental stage of height growth of plants in the layered forests (although the shrub species do not grow in height even at well-developed stages and have less effect on the varying in community types as compared to the tree species), are used to classify communities. Nevertheless, a maximum value of cover among the stratum alone is selected for each species and then used to predict the forest types of the future, even which are possibly affected by the minor values of other stratum. In many cases, this means that only the cover of the tall tree layer is used as the representative value for that species at the study site (Table 1).
Plot α
Layer/ Stratum
Plot β
By C-max procedure (species/cover rank)
By C-all procedure (species/cover rank)
By C-all procedure (species/cover rank)
By C-max procedure (species/cover rank)
Max cover Species A /3 Species B /3 Species C /1
Species A /1 Species B /3
I (tall tree)
Species A /3
Max cover Species A /3 Species B /3 Species C /1
II (sub-tree)
Species B /2
Species C /1
III (shrub)
Species C /+ Species B /1
Species A /3 Species B /1 Species C/1
IV (herbaceous)
Species B /3 Species A /1 Species C/1
Table 1.
Treatment of vegetation data in the respective analyses of procedures C-all and C-max. The Braun-Blanquet method records the occurrence species and cover in each layer/stratum separately. In the conventional C-all method, one representative value (i.e. the maximum value) of cover rank/coverage for each layer is extracted and used in the analysis, as shown in the table. In the C-all method introduced here, the same species appearing in different layers are treated as different species by attaching a symbol indicating each layer (e.g. I, II) to the species name.
We call this maximum coverage (over the layers) selected procedure as ‘C-max’. The conventional C-max procedure is fine on the community classification at the stable phase of community dynamics. However, that does not aim to survey the dynamics of the community at patchy/short-term scales, so it is inconvenient to recognize the community shape in the past and future in a changing environment. In reality, many plant communities are unstable and at a certain phase of those dynamics, especially under natural disturbances and human interferences.
3. Unselected uses of coverage (C-all procedure) and advantages
Here, we introduce the ameliorated procedures using data from the traditional Braun-Blanquet methods on the coverage of plants collected from the layered forests to consider the vegetation shifts such as forest regeneration and succession.
Imagine that a land is cleared in a temperate region. After a few decades, pioneer trees such as red pine Pinus densiflora would occupy forest overstory. However, beneath the overstory, the pine seedlings could not grow as a next generation of overstory because of the shading from their mother trees. Then the pine forest would shift to an oak forest dominated by Quercus serrata that is with much shade tolerance and grow up to the overstory from the saplings and seedlings at the shrub layer and the herbaceous layer. In phytosociological vegetation surveys, the presence of species is recorded separately for tree, sub-tree, shrub, and herbaceous layers. If all pieces of information from the vertical distribution of plants are used, it is possible to estimate such successional change (Table 1).
Therefore, we conduct every coverage of all layers selected procedure (C-all procedure) and show useful results to infer successional change drawn by the multivariate analyses (cluster analysis, INSPAN, and TWINSPAN) for community recognition. How to handle stratified data and the results of found possible vegetation shift are presented as below.
3.1 Handling the values of cover in the C-all procedure
As mentioned above, there are six cover ranks, and the median of their cover ranges was used in the following analysis (similar to the cluster analysis of Goto and Shimano [3], and Oyake et al. [4, 5]).
For the C-all procedure, when a species occurs in more than one stratum, cover values are treated separately for each stratum. For example, tall tree stratum = I, sub-tree stratum = II, shrub stratum = III, and herbaceous stratum = IV are recognized in a survey plot, and the cover rank of Species A is observed as
5 in the tall tree layer
no occurrence in the sub-tall-tree layer
3 in the shrub layer
1 in the herbaceous layer,
then treated as
Species A I is 5
Species A II is no occurrence
Species A III is 3
Species A IV is 1,
respectively. Such handling seems to be only natural but note that it is different from the conventional handling method, C-max procedure (Table 1).
3.2 Evaluation of vegetation status on the artificial green
The research group of Oyake, one of the authors, investigated the established vegetation on the embankment slopes of Japanese expressways approximately 50 years after construction. Such artificial greens are often needed to evaluate vegetation shifts to manage the lands, and the expected C-all procedure is useful for the evaluation.
Seven 5-m2 study plots with different stand of the sites were established along the Meishin Expressway (Shiga Prefecture, Japan [4]), four plots on the Kyushu Expressway (Kumamoto Pref., Japan), and four plots on the Chuo Expressway (Aichi and Gifu Prefectures, Japan. [5]). For each plot, established vegetation was recorded using the Braun-Blanquet method. When conducting the vegetation survey, the stratification method was used to record the occurrence species name and cover by species in each stratified layer (≥10 m: 1st (canopy) layer, 5–10 m: 2nd (sub-canopy) layer, 1.3–5 m: 3rd (lower tree) layer, 0.5–1.3 m: 4th (shrub) layer, <0.5 m: 5th (understory) layer.
The number of species and dominant species found in each plot are shown in Table 2. The vegetation at each plot was characterized as deciduous broadleaf forests (M2, M4, M6, M7, C1, C2, and C3 plots); plot C4 was Japanese red pine Pinus densiflora-dominated forest at an earlier successional stage; the plots M1, K1, and K2 were dwarf-bamboo communities dominated by Pleioblastus simonii; the plot M2 was dominated by herbaceous vine Pueraria lobata; the plot K3 was a community dominated by a giant bamboo Phyllostachys edulis, and the plots M5 and K4 were invaded by a giant bamboo Phyllostachys reticulata.
Deciduous broadleaf tree, Evergreen broadleaf tree
C3
44
Clethra barbinervis, Quercus glauca
Deciduous broadleaf tree, Evergreen broadleaf tree
C4
45
Pinus densiflora, Cerasus jamasakura
Evergreen conifer, Deciduous broadleaf tree
Table 2.
Number of species and dominant species at each survey artificial slope of expressway [4, 5].
A stratified cluster analysis was performed based on the results of the stratified vegetation survey obtained for the three routes. When the same species were recorded in different strata, two patterns of analysis were conducted: C-max and C-all procedures. Based on the clusters obtained, Indicator Species Analysis (INSPAN, see Box 1) was performed.
Cluster analysis summarizes groups of sites with similarities in species and displays a dendrogram. However, the dendrogram does not tell us which species contributed to that grouping. Here, INSPAN (Indicator Species Analysis) can be used to find representative species, which are significantly selected for each community grouping. It should be noted, however, that the indicator species selected will vary by grouping. For example, if one community is divided into two, the indicator species of the undivided community will be the indicator species of one of the two communities, and the other indicator species will be selected for the other community. Thus, it is important to determine how many groups are divided. In mentioned in Figure 3 is useful in determining the number of divisions. In general, a higher value of the “gap statistic” indicates an appropriate number of groups.
Box 1.
Cluster analysis and INSPAN.
3.2.1 Applications to cluster analysis and INSPAN
The results of cluster analysis based on C-all procedures and the indicator species indicated by INSPAN are shown in Figure 1. The results of cluster analysis by the conventional C-max procedure and the indicator species indicated by INSPAN are shown in Figure 2.
Figure 1.
Dendrogram of cluster analysis results and indicator species for each cluster from INSPAN by C-all procedure.
Figure 2.
Dendrogram of cluster analysis results and indicator species for each cluster from INSPAN by C-max procedure.
Figure 3.
Schematic diagram of how the number of cluster divisions is determined using the gap statistic. Usually, the highest value of gap statistic, the number of division, is optimum. Characters from A to I are community groups in the cluster. Here, six is the number of optimum division. Note that the number 6 has nothing to do with the value of height (dissimilarity) in the cluster analysis. The position (height) of the horizontal line separating the groups in the cluster diagram can be anywhere as long as the number of group divisions is six.
Looking at the results of INSPAN, it appears that dominant species have a significant influence on cluster partitioning in cluster analysis using the conventional C-max procedure (Figure 2). On the other hand, cluster analysis using the C-all procedure can extract species that characterize the vegetation, although they are not highly covered, such as the understory vegetation.
For example, plot M2, which was dominated by P. lobata, was classified as cluster 5 in the stratified cluster analysis, where many deciduous broadleaf forest plots were classified. In INSPAN, these understory/floor vegetations are shared with other deciduous broadleaf forest plots. Thus, the C-all procedure can detect similarities in communities that are difficult to discern from the dominant species. On the other hand, when split into eight clusters, plot M2 is classified in a different cluster from the deciduous broadleaf forest plots (see Box 1).
Thus, cluster analysis using the C-all procedure and INSPAN can be useful to predict future vegetation shifts.
3.3 Evaluation of the invading process of alien species to a natural vegetation
As a next example, we conducted a survey in a riparian forest in central Japan to obtain data, which is suitable to evaluate the vegetation dynamics, fearing that native Japanese willow Salix serissaefolia forest, which is unique to Japanese riverbanks and might be replaced by invaded black locust Robinia pseudoacacia. Based on this data, we show that two-way indicator species analysis (TWINSPAN, see Box 2) followed by the C-all procedure is suitable for such vegetation shift.
TWINSPAN is one of the standard analyses on vegetation classification. That is a top-down community classification method, similar to phytosociological tabulation, in which all sites in the study area are firstly divided into two groups based on species composition, and then the divided groups are further divided into two groups (Figure 4). Another representative method of classifying communities by species composition is cluster analysis, which is also used in [3], and is a bottom-up method in which sites with similar species compositions are grouped together based on the similarity of their compositions. Incidentally, Mineda et al. [6] pointed out that cluster analysis is appropriate when the species composition among sites is discontinuous and intermittent, while TWINSPAN is appropriate when it is continuous [6]. Here, we show the differences between the cluster analysis and TWINSPAN, and the output results from common software. Suppose we have data on plant species and their dominance for 100 stands (plots). In the case of cluster analysis, the occurrence of plant species at each site is evaluated numerically to determine how similar the occurrence of each species is and how similar the amount of occurrence is between the stands. TWINSPAN, on the other hand, overviews all plant communities of interest and divides them into broader communities; first there are two groups. After that, the groups will be divided into groups (Table 3).
Box 2.
TWINSPAN.
Figure 4.
Schematic diagram of the result of site group division with TWINSPAN. This method divide sites into two groups first, then the groups also do into two in the next step secondly. Characters from A to D mean the divided groups.
Species with layer
Plot number
Species division
3
2
8
4
9
7
6
1
5
7
Tree Species AI
5
4
5
5
3
0
Tree Species AII
1
2
1
1
2
0
Shrub Species CIII
1
+
1
+
+
0
Herb Species DIII
+
+
+
+
+
0
Herb Species DIV
+
0
Tree Species BI
5
4
3
5
4
1
Tree Species BII
1
2
3
1
2
1
Tree Species BIII
1
2
1
1
1
Tree Species BIV
1
+
1
+
+
+
1
Shrub Species CIII
+
+
1
Plot division
0
0
0
0
0
1
1
1
1
1
Table 3.
Image of a result of TWINSPAN. Using the occurrence similarity, all species with the layer and all sites were divided into twice. The image here shows only the first two divisions of species and plots. Roman numerical, I, II, III and IV, means tree, subtree, shrub and herb layer, respectively. As you will find, tree species B can be found beneath the overstory of tree species A, but tree species A plants was not beneath tree species B layer. This indicates the forest dominated by tree species A will be the forest dominated by tree species B in the future.
The data were obtained from a Braun-Blanquet’s phytosociological vegetation survey of 107 plots in the summer of 2008 along the banks of the Saigawa River (Matsumoto and Azumino cities, Nagano Prefecture), central Japan, to determine the species composition and characteristics of plant communities [3]. At each plot, the cover rank and community height of all species in each layer including the tall-tree layer (10 m ≤ tree height), sub-tall-tree layer (5 m ≤ <10 m), shrub layer (0.7 m ≤ <5 m), and herbaceous layer (<0.7 m) were recorded.
In the study site, several community phases along an invasion intensity of R. pseudoacacia populations into S. serissaefolia stand were observed. Data were collected by the method of Braun-Blanquet.
3.3.1 Application to TWINSPAN
In the present study, TWINSPAN was conducted using values compiled by the C-max and C-all procedures to determine the likelihood of coexistence of species groups that characterize the community within a grouped community (plot group), and the sympatry and exclusivity within and among species between the upper and lower layer [7]. The analysis was performed using the “twinspan” function of the “twinspanR” package (ver. 0.19) on the statistical analysis software R (ver. 3.5.0; R Core Team [8]). Here, only the result with C-all procedure was shown.
The TWINSPAN results originated from [7] allowed us to recognize the communities as four major groups. The first two divisions from the top of the dendrogram were herb/shrub and forest communities. In the next division, “the herb/shrub community” was divided into R. pseudoacacia community and S. serissaefolia shrub/grass community. On the other hand, “the tall tree community”’ was divided into R. pseudoacacia and S. serissaefolia communities. However, it is noteworthy that R. pseudoacacia forests had R. pseudoacacia shrubs and seedlings, while S. serissaefolia forests did not have S. serissaefolia seedlings or shrubs, but R. pseudoacacia seedlings and shrubs (Figure 5). Although the TWINSPAN analysis can be divided into further small groups, no further dividing was necessary to determine the vegetation shift betweenR. pseudoacacia and S. serissaefolia communities.
Figure 5.
Summarized result of plot groups with TWINSPAN. Community types and key species are described.
That is, R. pseudoacacia IV, which is a seedling or juvenile tree, grows both beneath the forest overstory of S. serissaefolia I and R. pseudoacacia I, whereas S. serissaefolia IV seedlings would not grow under the canopy of R. pseudoacacia. S. serissaefolia IV mainly occurred in sandy bare soil sites. This asymmetric distribution of R. pseudoacacia seedlings and juvenile trees allowed them to invade S. serissaefolia tall forests, while S. serissaefolia seedlings and juvenile trees could not invade either S. serissaefolia forests or R. pseudoacacia forests. These suggest that, in the future, S. serissaefolia willow forests may be replaced by R. pseudoacacia forests, but R. pseudoacacia forests will remain as R. pseudoacacia forests.
Thus, in contrast to the conventional C-max procedure, TWINSPAN based on the C-all procedure made it possible to infer what species will replace and also diagnose future changes in dominant species. Combining further studies on variations in ecophysiological traits, population dynamics, and relationships among species would be nice with long monitoring of vegetation change.
In the classical phytosociological vegetation surveys, the values of plant coverage are obtained for layered communities. In the conventional classification and/or recognition of vegetation, only the maximum coverage over the layers is selected (C-max procedure). We use our original C-all procedure (every coverage of all layers selected procedure) and new success to reveal the vegetation shift such as successional processes from the phytosociological data. The C-all procedure will be useful to not only the layered forests but also the herbaceous communities and grasslands (see [5]).
Other dominant indices such as diameter and height of trees would be able to apply similarly other than cover value. However, general tree-by-tree surveys often miss information on the plant layer with less than 1.3 m in height. The manner of phytosociological surveys, which basically intends lower layer(s), and C-all procedure complements such problems easily. Re-analyzing vastly accumulated cover data from the past would give additional valuable information on vegetation history by C-all procedure.
Recently, new informatics to measure vegetation structure has been developed, and the data from those could be also applied by the C-all procedure in the future. However, the convenient cover data by the traditional phytosociological surveys should continue to use important information in many fields.
We, the authors, would like to thank our editor, who invited us to publish this book and was patient with our late submission of the manuscript.
References
1.Braun-Blanquet J. Pflanzensoziologie, Grundzüge der Vegetationskunde. 3rd ed. Berlin: Springer-Verlag; 1964
2.Mueller-Dombois D, Ellenberg H. Aims and Methods of Vegetation Ecology. New Jersey: The Blackburn Press; 1974
3.Goto S, Shimano K. Riparian forest invasion by Robinia pseudoacacia and its effects of riverside vegetation. Vegetation Science. 2018;35:49-65
4.Oyake Y, Imanishi J, Ishihara K, Ogura I, Shibata S. Long-term vegetation transition on man-made slopes 53 years after construction in Central Japan. Landscape and Ecological Engineering. 2019;15:363-378
5.Oyake Y, Mimaki R, Oda K. Vegetation on artificial embankment slopes of the expressways approximate 50 years after construction in two express ways: Kyushu and Chuo Expressway. Journal of the Japanese Society of Revegetation Technology. 2023;48:507-515 (in Japanese with English Abstract)
6.Mineta T, Yamanaka T, Hamasaki K. Statistical analysis for biological and social research (8): Classification (cluster analysis, and indicator species analysis). Journal of the Japanese Society of Irrigation, Drainage and Rural Engineering. 2005:221-226 (in Japanese)
7.Shimano K, Goto S, Kobayashi T. Shifts from native to non-native riparian plant communities analyzed using hierarchical phytosociological data. Japanese Journal of Ecology. 2022;72:13-25 (in Japanese with English Synopsys)
8.R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2018
Written By
Koji Shimano, Yui Oyake and Tsuyoshi Kobayashi
Submitted: 29 October 2023Reviewed: 29 October 2023Published: 13 December 2023
Open access peer-reviewed chapter
