The density and the age distributions of E. polyandra, a typical pioneer ... Accepted 4 August 1992. .... on the east side of the landslide scar, and (4) the slope on ...
Ecological Research (1993) 8, 4 7 - 5 6
Vegetation pattern and microtopography on a landslide scar of Mt Kiyosumi, central Japan AKIKOSAKAIAND MASAHIKOOHSAWA
Laboratory of Ecology, Faculty of Science, Chiba University, Yayoi-cho, Inage-ku, Chiba, 263Japan
Vegetation pattern and microtopography were examined on a mountain slope with a rotational type landslide scar on Mt Kiyosumi, central Japan. Similarities of distribution patterns among 55 woody species were calculated using Cole's species association coefficient, and based on them, seven vegetation units were classified using cluster analysis and principal coordinates analysis. The seven vegetation units coincide with seven microtopographical facets a t 101 t o 1 0 2 m 2 order. Furthermore, these vegetation units were grouped into three higher categories by reciprocal averaging and principal coordinates analysis. They were ridge slopes, surrounding slopes and landslide slopes. The three categories were arranged in the above-mentioned order based on similarity in floristic composition. In the ridge slopes, late-successional trees and deciduous trees had high relative basal areas. In the surrounding slopes, Euptelea polyandra and other deciduous trees had high relative basal areas. In the landslide slopes, E. polyandra and deciduous shrubs had high relative basal areas. The density and the age distributions of E. polyandra, a typical pioneer tree which invades disturbed sites, suggested that the severity of soil surface disturbances increase in this order. The disturbance regime explains the vegetation pattern on the study site, where the rotational type landslide had occurred. Key words: disturbance regime; Euptelea polyandra Sieb. et Zucc.; landslide; microtopography; vegetation pattern analysis.
INTRODUCTION In mountainous regions with heavy rainfall, landslides are a common disturbance agent that play an important role in determining vegetation structure and species composition (Veblen & Ashton 1978; Garwood et al. 1979; Veblen et al. 1980; Shimokawa & Jitousono 1984; Nakamura 1990). After the occurrence of landslides, pioneer herbs and trees opportunistically invade the landslide scars, and construct vegetation communities different from those of surrounding forests (Langenheim 1956; Guariguata 1990). It is also known that different types of vegetation develop even within single landslides corresponding to the microhabitat conditions (Langenheim 1956; Flaccus 1959; Sakura & N u m a t a 1973; Guariguata 1990). Previous authors have noted that microtopography is an important factor affecting the vegetation patterns
Accepted 4 August 1992.
and establishment of plants in such disturbed habitats. However, the relationships between vegetation patterns and microtopography in landslide scars are not yet fully explored. In the present paper, we statistically analyze the vegetation pattern in relation to the microtopography in and around a landslide on a mountain slope, and discuss the causal mechanisms determining the characteristic vegetation pattern in and around the landslide.
STUDY AREA The study area is located in a warm-temperate forest of the Tokyo University Forest on Mt Kiyosumi, Chiba, central Japan (Fig. 1). There are many landslide scars due to the steep topography of loose Tertiary deposits (Iijima & Ikeya 1976) and the high annual rainfall of 2232 m m per year from 1975 to 1984 (Tokyo University Forest in Chiba, 1987). One of the largest landslide scars was selected for this study (1254 m2). The site is located on a
48
A. Sakai and M. Ohsawa
',',':l I'
Chiba
p
~3~_~~i
t 7~~ r'~l~ -~Jl
fi/C
Ocean 35~
140~ Contour is 200m art. I
0
i
I
100
i
_A
N
200m
Fig. 1. Thelocationofthestudysite.Theboldlinesintherightfigurearerivers. north-facing slope, ca 200 m in altitude, along the Maesawa River, one of the small tributaries of the Obitsu River (Fig. 1). The vegetation of this region consists of natural forests, coppice forests and manmade conifer plantations. The natural forests and the coppice forests are dominated by late-successional evergreen broadleaved trees such as Castanopsis cuspidata var. sieboldii, Quercus acuta and Quercus salicina, and conifers, Abiesfirma and Tsuga sieboldii. The large landslide studied is located in the coppice forest tract no. 29, which was logged 43 years ago (Tokyo University Forestin Chiba, 1986).
METHODS Data sampling Microtopography within the study area was measured and a contour map was made in 1989 to a scale of 1:100. Microtopographical elements such as breaks of slopes and cliffs were recorded in detail on this map. Soil depths were measured at 2 m • 2 m meshes that cover the whole area of the study site in
1990. In 1989, we mapped the locations of all trees and shrubs taller than 1 m, and measured their diameters at a height of 1 m. Basal areas were calculated for all stems from their diameters.
Vegetation pattern analysis To analyze the vegetation pattern, we tried to divide the study site into vegetational units using statistical methods. The study site was divided into 2123 meshes of 2 m • 2 m. Here, in order to reduce sampling errors, a mesh was overlapped with its eight neighboring meshes. That is, the centers of four neighboring meshes were located at the four angles of the mesh concerned, and the centers of the other four neighboring meshes were located at the middle points of the four sides. Presence or absence of each species in each mesh was recorded. Cole's (1949, 1957) species association coefficient C was calculated using the presence or absence data for all pairs of 55 species occurring five or more times in the study site, to evaluate the similarities of distribution pattern of the species. Values of Cole's C range from - 1 to + 1. If the two species concerned occur independently, the C value is 0. If they tend to occur
Vegetation of a landslide together, the C value is positive. If they tend to occur separately, the C value is negative. The Cole's C values were transformed into the values of a dissimilarity index (x) ranging from 0 to + 1 for the cluster analysis as follows: x = i - (c + I)/2
(I)
The x's were adopted in the duster analysis using the average linkage method, and the species were classified into species groups of similar distribution patterns. Cole's C was recalculated for each of all pairs of the groups and principal coordinates analysis (Gower 1966), which was a kind of multivariate analysis, was conducted using these C values. Sufficient ratio of cumulative contribution was obtained at the third dimension (84%). Hence, the relative similarities of the distribution pattern among the species groups were expressed by the distances of the species groups in three dimensions. The values of each of the three axes were calculated for all the meshes having one or more plants as follows. If all plants in the mesh concerned belonged to the same species group, the three values of the mesh were defined as the three values of the species group. If a mesh had more than one plant bdonging to different species groups, the three values of the mesh were defined as the averages of them. Vegetational units were recognized from the three maps showing the distribution patterns of values on the three axes. In this paper, we refer to these vegetational units as vegetation units. Vegetation analyses were conducted for the vegetation units extracted. To ordinate the vegetation units, reciprocal averaging (Hill 1973) was conducted using relative basal areas of each species. Shannon's species diversity index H ' (Pielou 1977) of each vegetation unit was calculated as follows:
H' = - • Pi log2 Pi
(2)
where Pi is the relative basal area of the ith species.
Disturbance history
49
of 11 individuals were sampled at a height of 30 cm. Figure 2 shows the relationship between the diameters at a height of 1 m and the ages at a height of 30 cm. The ages of the other E. polyandra stems were estimated from the diameters using this relationship. If an individual had multiple stems, its age was determined from the oldest stem.
RESULTS
Microtopography and soil c o n d i t i o n Figure 3 shows a contour map (a), geomorphological explanation (b) and soil depths (c) of the study site. There was a large, clear landslide scar, which consisted of several small landslide scars, in the center of the study site (Fig. 3a). The microtopography of the study site was heterogeneous, and there were many micro-scale topographical facets bordered by cliffs and breaks or changes of slopes (Fig. 3b). In the landslide scar, soil depth was nearly zero and bedrock was weathered in some parts, while in the surrounding sites, soil depth was relatively thick (Fig. 3c).
Vegetation patterns There were 89 woody species including 31 deciduous trees, 15 evergreen broad-leaved trees, 6 conifers, 26 deciduous shrubs, 3 evergreen shrubs, and 8 climbers in the study site. O f these, the 5 5 species
30
9
' / oE
20
o
g
9
9
9
9
9
9
,.5
10
