全 文 ：利用影像判读与群落监测分析长白山针叶林动态 3
刘琪　 3 3 　李轩然　胡理乐
(中国科学院地理科学与资源研究所 ,北京 100101)
【摘要】　基于不同空间尺度即永久标准地的群落学调查与卫星遥感监测相结合研究了长白山亚高山针叶
林的结构与动态. 标准地两次调查的间隔为 11 年. 结果表明 ,每 10 年的死亡率为 7 %～9 % ,进界比率为
18 %～20 %. 鱼鳞云杉、臭冷杉及岳桦可以在林冠下顺利完成更新 ,而长白落叶松为先锋种 ,只能在林窗或
裸地更新. 落叶松为云杉及冷杉提供良好的更新条件 ,从而在维持亚高山森林的稳定性上起着重要作用.
成熟林密度 (1 000 株·hm - 2左右)变化不大. 利用美国陆地卫星 TM 图像分析 1984～1997 年植被变化表
明 ,大规模的风倒等自然干扰很容易检测出来. 从景观尺度上 ,利用 TM 图像监测植被变化非常有效. 不同
反射强度变化的象素数量统计结果表明 ,群落的进展演替与逆向演替同时存在 ,并处于相对平衡状态. 但
由于图像分辨率 (30 m ×30 m)的关系 ,林窗很难同噪声区分开来. 长白山亚高山针叶林带因有大量的落
叶松斑块而呈现出镶嵌结构 ,这种镶嵌结构被定义为亚高山植被的顶极状态. 在小尺度上 ,例如面积仅为
几公顷的标准地 ,特别是混有落叶松等先锋树种的群落 ,种类组成随时间而变化 ,但是卫星图像分析结果
证明 :整体上 ,亚高山植被处于稳定状态.
关键词 　针叶林 　变化检测 　森林动态 　陆地卫星 TM 　树木枯死
文章编号 　1001 - 9332 (2004) 07 - 1113 - 08 　中图分类号 　S77118 ;S75712 　文献标识码 　A
Image analysis and community monitoring on coniferous forest dynamics in Changbai Mountain. L IU Qijing ,L I
Xuanran and HU Lile ( Institute of Geography and N atural Resources , Chinese Academy of Sciences , Beijing
100101 , China) . 2Chin. J . A ppl . Ecol . ,2004 ,15 (7) :1113～1120.
The structure and dynamics of coniferous forests in Changbai Mountain were studied at different spatial scales ,in2
cluding ground survey of permanent plots and analysis of multitemporal satellite images. Plot2scale examinations
showed that the mortality rate was 7 %～9 % ,and the recruitment rate was 18 %～20 % per 10 years. Species
composition changed over time. Picea jezoensis var. microsperm a , A bies nephrolepis and Betula erm anii pre2
sented a self2maintaining capability ,because they could regenerate under canopy. L arix olgensis was a pioneer
species and could regenerate only in open land or gaps. This species played an important role by providing condi2
tions for the regeneration of spruce and fir. The tree density in the mature forest was 1 000 stems·hm - 2 for trees
bigger than 3 cm in diameter ,which showed no significant variations among different stands. Landsat TM images
were used for detecting the cover changes from 1984 to 19971Large scales of wind throw were detected by this
approach. Based on the analysis of radiance changes at the landscape scale ,the pixel number of the disturbed area
was similar to that of the succeeding stands ,suggesting that the forest was in a state of equilibrium. Fine gaps ,
however ,were difficult to identify with the TM data because of its coarse resolution. The mosaic structure of the
subalpine vegetation was characterized by scattered larch patches. At the landscape level ,the vegetation was in a
stable stage.
Key words 　Coniferous forest , Change detection , Forest dynamics , Landsat TM , Mortality.3 The study was supported by the National Key Basic Research Special
Foundation of China (No. 2002CB4125) .3 3 Corresponding author.
Received 2003 - 06 - 23 ,Accepted 2003 - 11 - 05.
1 　INTROD UCTION
　Coniferous forest is widely dist ributed in subalpine
of temperate regions , especially in northeastern Asia
and the northern area near the arctic[13 ,14 ] . This for2
est type is generally considered as a climax[26 ] . Within
the Biosphere Reserve of Changbai Mountain in Chi2
na ,coniferous forest accounts for nearly a half of the
whole reserve in area[13 ] . It is primarily dist ributed
from altitude 1 100 m to 1 700 m on Changbai Moun2
tain[11 ,27 ] . The composition and succession of this for2 est have been reported in many studies during the pastfour decades[2 ,8 ,27 ,31 ] . The establishment and dynam2ics of the forest were also reported based on regenera2tion pattern and age structure[10 ,17 ] . Still ,the functionof larch in this forest remains questioned , i. e. willlarch eventually be excluded from the subalpine ordoes it play an important role in maintaining the com2munity stability ? To better understand the role of
应 用 生 态 学 报 　2004 年 7 月 　第 15 卷 　第 7 期 　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　
CHIN ESE JOURNAL OF APPL IED ECOLO GY ,J ul. 2004 ,15 (7)∶1113～1120
larch in landscape scale ,studies need to be conducted
over a long2time period ,which is practically impossi2
ble for one to do such observations.
　Plant ecologists generally predict the succession
trend of a community by data collected at a certain
time point [3 ,4 ,11 ] . However , the plot f rom which in2
formation is extracted represents only a tiny unit of
the entire community , since it shows a succession
stage for just that patch alone. To analyze the stability
of a community , the species alternation and recruit2
ment pattern are usually the factors that are consid2
ered[17 ,22 ] . Predictions of community changes based
on a small sample area are often debatable. The scale
of the study object is often ignored but which is a
critical issue for describing community dynamics.
Therefore ,it is questionable to represent the dynamics
of a forest community by the successional t rend of a
plot which covers merely one or several hectares. The
small area of forest community we refer to is just a
point versus the whole ecosystem. Such points may
show dramatic changes over time ,while at the land2
scape level , the phase of an ecosystem may stay un2
changed. To demonstrate equilibrium or dynamics ,
field investigation is important . But it is almost im2
possible to conduct long2term field investigations due
to high cost involved in this kind of studies in addition
to the limitations of human power.
　Satellite remote sensing provides a meaningful
method for detecting vegetation and land cover
changes[5 ,7 ,21 ] . Recently , the monitoring of forest
change under various disturbances has been intensive2
ly conducted thanks to the accumulation of satellite
data[15 ,18 ,23 ] . This technique can be used for analyzing
the changes of the entire ecosystem. By comparing the
images taken at different times , landscape2scale
changes can be detected[6 ] . The main methods are im2
age difference ,principal component analysis[16 ,24 ] ,and
post2classification difference[1 ,12 ] .
　This paper is aimed to clarify the stability and in2
stability of forest community in different scales and
their unity under natural conditions. In particular ,the
effectiveness of using satellite data to describe the sta2
bility of zonal vegetation in macro scale is discussed.
The dynamics of different populations on the basis of
size st ructure and its changes are examined. For land2
scape scale ,image differencing is applied to extract the
change of coniferous forest in the subalpine of Chang2
bai Mountain.
2 　METHODS
　 In 1981 , Changbai Mountain Forest Ecosystem
Station , The Chinese Academy of Sciences set up two
permanent plots ,plot 2 and plot 3 ,in the coniferous
forest zone on the north slope of Changbai Mountain.
Plot 2 was located at 128 :07 :53E/ 42 :08 :38N ,1 260
m above sea level (asl) ,and plot 3 at 128 : 03 : 54E/
42 :04 :36N ,680 m asl. The size of each plot was one
hectare. Trees with DBH > 8 cm were recorded ,and
crown projection was mapped. In 1992 ,the two plots
were re2measured. At this time , t rees with DBH > 3
cm were recorded. During the second survey ,vegeta2
tion on the forest floor was investigated by recording
the dominance ,coverage and height . In 1989 ,a tem2
porary plot (plot 8921) of 50 m by 20 m at 128 :08 :
15E/ 42 : 07 : 55N , 1 300 m asl was measured. This
plot was dominated by larch. The topography of the
study area was relatively gentle , with its slope less
than 10 degrees. These plots are named with domi2
nant species : Korean pine2spruce2fir forest for plot 2 ,
mountain birch2spruce2fir forest for plot 3 , and
spruce2fir2larch forest for plot 8921 ,respectively.
　For detecting the disturbance of large scales like
wind throw , two cloud2f ree Landsat TM ( band 1)
images (Nov. 18 ,1984 and Nov. 7 ,1997) were used
for gap identification. The images were co2registered
using the fourth order polynomial method. About 100
control points were picked from each image. The error
(RMS , root mean square) was controlled within one
pixel. By image differencing , the changes of forest
cover during 1984～1997 was confirmed. Before im2
age differencing , relative atmospheric correction was
performed. The 1984 image was used as a reference
image ,and the 1997 image was corrected by using a
lineal regression function. For which ,objects like con2
crete and water ,which are constant in reflectance re2
gardless atmospheric condition , were used for estab2
lishing regressive function. The number of pixels with
different changes were counted for the entire conifer2
4111 应 　用 　生 　态 　学 　报 　　　　　　　　　　　　　　　　　　15 卷
ous forest zone. The radiance was represented by digi2
tal number (DN) .
3 　RESULTS
311 　General characteristics of forest stands
　Korean pine2spruce2fir forest (plot 2 ) was dis2
t ributed near the deciduous broad2leaved2coniferous
mixed forest zone ,showing a transitional feature with
trees of Pi nus koraiensis . The density of this plot was
777 stems·hm - 2 ,basal area was 4818 m2·hm - 2 ,and
average DBH was 2419 cm. The maximum DBH was
6012 cm ( Pi nus sylvest rif orm is) . A total of eleven
tree species were identified in this plot . The dominant
species ,which were defined as > 10 % of total basal
area , include Pi nus koraiensis ( 2413 %) , Picea je2
zoensis var. kom arovii ( 2411 %) , Picea koraiensis
(1218 %) , and A bies nephrolepis ( 2114 %) . L ari x
olgensis accounted for only 617 %. Broad2leaved
species were rare. The vertical profile of the commu2
nity , with a high closure , was continuous , and the
dominant t ree layer was occupied by Pi nus koraien2
sis , Picea jezoensis var. kom arovii and Picea koraien2
sis. The subcanopy layer was dominated by A bies ,
which had the maximum DBH of 2312 cm. The
trunks were densely covered by mosses ,indicating wet
and dark environment . This condition heavily rest ricts
the species richness in the undergrowth. Lichens had
little dominance. Seedlings of A bies nephrolepis and
Picea jezoensis var. kom arovii were dense. The
sapling ( > 115 m in height and < 8 cm in diameter)
layer was dominated by A bies alone.
　The mountain birch2spruce2fir forest (plot 3) was
dominated by Picea jezoensis var. kom arovii
(4119 %) . A bies accounted for 25 % ,and Bet ula er2
m anii 1516 %. This plot had 7 species. Stand density
was 1134 stems ·hm - 2 . The total basal area was
43125 m2·hm - 2 ,and mean DBH was 1819 cm. The
maximum DBH was 65 cm ( Picea) . Pi nus and Picea
koraiensis were absent in this plot . Instead , Bet ula
erm anii was the co2dominant species. The vegetation
of the undergrowth was similar to plot 2 , but the
crown density was slightly lower than.
　The spruce2fir2larch forest (plot 8921) ( Fig. 1)
was a secondary stand which was in its building
phase. The stand density was 980 stems·hm - 2 , and
basal area was 5913 m2·hm - 2 . L ari x was dominant
in this plot , and its relative basal area was 7913 % ,
while its relative density was only 1514 %. There was
absent of young individuals of L ari x . A bies presented
a relative density of 3815 % ,and Picea jezoensis var.
k om arovii 919 %.
Fig. 1 DBH distribution of tree populations in larch2spruce2fir
forest (1 300 m asl) .
312 　Dynamics of t ree populations
　The density of plot 2 increased from 690 to 777
stems·hm - 2 during the past decade. The average
DBH decreased from 2716 to 2419 cm. This corre2
sponded with the increase of young individuals of 8～
12 cm in diameter. The basal area stayed at the simi2
lar level ,4813 m2·hm - 2 in 1981 ,and 4818 m2·hm - 2
in 19921The crown density ( DBH > 8 cm ) was
6913 % ,of which 59 % was evergreen and 1013 %
was deciduous trees ( mainly L ari x ) , respectively
( Table 1) .
Table 1 Tree species composition and its changes in Korean pine2spruce2
f ir forest ( plot 2) during the period of experiment
Species D81 D92 RN81 RN92 RBA81 RBA92
Pinus koraiensis 3015 3019 01280 01247 01333 01338
Picea koraiensis 3318 3610 01171 01131 01235 01225
Picea jezoensis var. kom arovii 2315 2415 01302 01247 01214 01213
L arix olgensis 3110 3219 01082 01069 01096 01100
Pinus sylvest rif orm is 4217 4316 01024 01021 01051 01053
Betula platyphylla 2616 2718 01056 01034 01045 01034
A bies nephrolepis 1514 1015 01085 01197 01025 01031
Acer ukurunduense 916 + 01034 + 01004
Acer tegmentosum 918 + 01015 + 01002
Tilia am urensis 1210 + 01002 + +
Acer mono 915 + 01002 + +
D81 :Mean DBH (cm) in 1981 ;D92 :Mean DBH in 1992 ; RN81 : Relative densit y
in 1981 ; RN92 : Relative density in 1992 ; RBA81 : Relative basal area in 1981 ;
RBA92 :Relative basal area in 1992. The same below.
　The DBH dist ribution patterns in plot 2 were dif2
ferent among populations ( Fig. 2) . It can be roughly
divided into 2 types , reverse J2shape and normal dis2
t ribution. A bies was the only one that showed a re2
verse J2shape , which occurred mainly in the sub2
canopy layer. Stems of A bies larger than DBH 24 cm
51117 期 　　　　　　　　　　　刘琪　等 :利用影像判读与群落监测分析长白山针叶林动态 　　　　　　　　　　　
were very few. Pi nus koraiensis and the two Picea
species represented a normal dist ribution , t heir density
were decreasing toward both small and large DBH
classes. L ari x and Pi nus sylvest rif orm is were scat2
tered in large DBH classes and few saplings or young
trees were found under canopy. The average DBH of
the two populations were higher than that of most
other species. The DBH dist ribution of these two
species was irregular ,showing a degrading trend. This
result was very similar to previous report on this
area
[10 ]
. Such features of the population structure sug2
gested the differences in shade tolerance and regenera2
tion process ,as indicated by Saxena et al. [22 ] .
Fig. 2 Size distribution and its changes during 112year period in
pine2spruce2fir forest (1 260 m asl) .
　The highest increment of DBH was 713 cm ( Picea
j . , DBH = 1516 in 1981 cm) and the majority of
t rees had an increment less than 210 cm during the 11
years. The relative increment of most individuals was
less than 15 %. The increment showed no significant
correlation with DBH ,implying the st rong effects of
micro conditions on growth. Those having small in2
crement rates are considered the results of upper2
canopy oppression. It may also relate to the lateral in2
terference that a higher growth rate is positively cor2
related with the fast growing of the nearest t rees and
the distance from which[19 ] .
　A total of 70 stems·hm - 2 died or fell during the
112year interval. The mortality rate was 1011 %
(019 %·yr - 1) . The size dist ribution of dead individu2
als showed a pattern consistent with that of the stand2
ing trees ,a normal dist ribution with the peak at DBH
20～24 cm. The mortality ratios varied from 0 % to
30 % ( Table 2 ) . This was consistent to the DBH
structure. L ari x and Pi nus sylvest rif orm is had no
individuals under the dominant canopy , and showed
the lowest mortality ratios. Shade intolerance is con2
sidered the cause for those with a high mortality rate.
Bet ula is an example of a shade intolerant species ,
which was lack of seedlings under canopy. Natural
death and windfall are considered permanent condi2
tions affecting the recruitment2death balance[25 ] .
Table 2 Mortality and recruitment of Korean pine2spruce2f ir forest
during the period of experiment
Species DBH BA % MRT % REC
Picea koraiensis 3012 33114 1410
Picea jezoensis var. komarovii 2610 27140 818 2
Bet ula platyphylla 2512 14160 3014
A bies nephrolepis 2013 12130 2517 127
Pinus koraiensis 3210 8126 216 3
L ari x olgensis 2515 4129 519
Acer mono - 1
Tilia am urensis - 1
Pinus sylvest rif ormis -
Total 8143 1011 134
DBH : Mean diameter (cm) ,BA :Basal area , MRT % : Mortalilt y ratio ,
and REC :Recruited individuals (hm - 2) .
　There were 134 individual t rees recruited during
the 11 years , and the recruiting ratio was 1914 % ,
nearly 2 times of the mortality ratio . Most of the new2
ly recruited tree species were A bies , accounting for
95 % ( Table 2) . The saplings (DBH 3～8 cm) of t ree
species was A bies only ,and the density was as high as
2 697 stems·hm - 2 . This result agreed with the previ2
ous report [27 ] . A bies is a shade tolerant species , and
the densely crowded canopy layer made the niche sig2
nificantly dark ,and it was therefore difficult for other
species to colonize the same site.
　In brief ,regeneration patterns agreed with the hy2
pothesis that those having more individuals in small
DBH classes showed a property of continuous recruit2
ment [20 ] , like A bies , while those without young co2
horts ,like Picea and Pi nus etc. ,were discontinuous.
　Birch2spruce2fir forest (plot 3) showed a little dif2
ference from plot 21All of the three dominant species
( Picea , A bie , Bet ula) presented a equilibrium status ,
and the st ructure had no significant changes during
the 112year time ( Fig. 3) . The crown density of this
forest was lower than plot 2 , while the regeneration
rate of plot 3 was much higher than plot 21The densi2
6111 应 　用 　生 　态 　学 　报 　　　　　　　　　　　　　　　　　　15 卷
ty of adult t rees (DBH > 8 cm) increased from 1 088
stems·hm - 2 to 1 134 stems·hm - 2 . The total basal
area was 3810 m2·hm - 2 in 1981 and 4217 m2·hm - 2
in 1992 ,and the relative increment was 1213 %.
Fig. 3 Size distribution and its changes during 112year period in
birch2spruce2fir forest (1 680 m asl) .
　　The three dominant species presented a similar
dist ribution mode , but in different extents ( Fig. 3) .
The highest f requency of A bies occurred in the small2
est DBH classes ,and the size range was under 40 cm ,
meaning that most individuals were in the subcanopy.
The DBH range of Picea jezoensis var. kom arovii was
much wider (maximum size 68 cm) than others ,indi2
cating a high dominance in the canopy layer and a
strong shade tolerance. The enormous saplings (DBH
3～8 cm) (280 stems·hm - 2) of Picea jezoensis var.
kom arovii showed a great potential for regeneration
under shade environment . A cer ukurunduense was a
large2sized shrub species in the subcanopy. The DBH
was smaller than 20 cm.
　The DBH increment was considerably uneven ,
ranging from 011 to 910 cm. The increment of most
individuals was less than 410 cm. The increment ratio
was up to 50 % ,with most of the individuals showing
less than 30 %. This unevenness , especially in the
same DBH class ,was probably due to the differences
of vertical position in the community i . e. smaller for
the oppressed , and larger for the opposite. As a
whole ,the maximum relative increment of each DBH
class decreased with DBH size ,f rom 52 % to nearly 0
( Table 3) . This pattern of growth ratio showed the
crowding status of the trees that were negatively af2
fected by neighboring trees.
Table 3 Tree species composition and its changes in mountain birch2
spruce2f ir forest ( plot 3) during the period of experiment
Species D81 D92 RN81 RN92 RBA81 RBA92
Picea jezoensis var. kom arovii 2116 2214 01410 01420 01601 01611
A bies nephrolepis 1510 1610 01283 01257 01167 01159
Betula erm anii 1812 1916 01154 01159 01140 01152
Sorbus pohuashanensis 1413 1412 01094 01095 01046 01043
Acer ukurunduense 1116 1118 01051 01065 01016 01020
L arix olgensis 5316 5415 01004 01002 01024 01011
Pinus koraiensis 2817 3315 01002 01002 01003 01004
Picea koraiensis 2116 01002 + 01002 +
　There were 76 stems·hm - 2 died during the past 11
years , which were mainly the result of windthrow.
The death ratio was 710 %. Paralleled with the DBH
dist ribution of living trees ,the highest death ratio was
in DBH 8～ 12 cm , and it dropped sharply toward
larger sizes ( Fig. 3) . The death in basal area was 3122
m
2·hm - 2 (relative value 815 %) . Most of the dead in2
dividuals were A bies and Picea. L ari x had only 2 in2
dividuals in the 015 hm2 plot 11 years ago ,and one of
which snagged during the period. The death ratio was
high in those species without succeeding trees , like
Picea koraiensis and L ari x ,and low in the opposite
group ,like Picea jezoensis var. kom arovii . Subcanopy
species showed low mortality ratios ,like Sorbus and
A cer. Bet ula erm anii , which was abundant in the
stand to begin with and remained abundant after the
11 years. It showed a strong adaptability to the dark
conditions( Table 4) .
Table 4 Mortality and recruitment of mountain birch2spruce2f ir forest
during the period of experiment
Species DBH BA % MRT % REC
Picea jezoensis var. komarovii 2517 44172 4148 60
A bies nephrolepis 1513 31131 14129 32
L ari x olgensis 5511 14179 50100
Sorbus pohuashanensis 1912 5181 5188 14
Picea koraiensis 2116 2127 100100
Acer ukurunduense 1510 1110 3157 24
Bet ula erm ani 14
Total 1916 8148 6162 144
　The number of recruited trees was 144 stems ·
hm - 2 , making a recruitment ratio of 1312 %. This
figure was considerably larger than the mortality ratio
( 710 %) . Most of the individuals were A bies and
Picea , which agreed with the mortality pattern de2
scribed above. A tremendous number of saplings of the
canopy2layer species were established on the ground.
71117 期 　　　　　　　　　　　刘琪　等 :利用影像判读与群落监测分析长白山针叶林动态 　　　　　　　　　　　
The sapling layer was dominated by Picea jezoensis
var. kom arovii (280 stems·hm - 2 ) ,and A bies (178
stems·hm - 2 ) . Bet ula was the third most dominant
species (50 stems·hm - 2) .
313 　Natural disturbance detected by satellite images
　By comparing two Landsat TM images ,a large and
bright area on the western slope was visible ( Fig. 4) ,
which indicated the damage by a typhoon in
19861 Image differencing ( Fig. 5) shows that the de2
creased values of < - 102 (in digital number) repre2
sented the windfall area , and the proportion of the
open land accounted for 616 % of total area. The pre2
Fig. 4 TM images (band 1) of Nov. 19 ,1984 (left) and Nov. 7 ,
1997 (right) .
Note that there is a large wind throw on the western slope of
the 1997 image , which was created by a strong typhoon in
19861Scale = 10 km.
Fig. 5 Radiance change of TM band 1 in coniferous forest from
1984 to 19971
Horizontal axis : change of radiance in digital number. Minus :
Vegetation decreased ,Plus :Vegetation increased.
disturbace area was very dark in the 1984 image ,
while the post2disturbance area was very bright in the
1997 image. Fig. 5 also shows that the numbers of
pixels in restoration and degrading were nearly the
same. This reveals the stability of the forest communi2
ty which was characterized by the mosaic st ructure of
mature and secondary patches. In other words ,the in2
crease of radiance means that the stand was reaching a
mature status ,while the decrease represents the for2
mation of a secondary patch.
　Large open areas ( > 10 hm2 ) created by wind
throws were very different f rom relatively small one.
They were thoroughly cleared of saplings or large
sized trees. Moreover , according to the dendrology
study in this area[28 ] ,t rees that regenerate f rom seeds
or seedlings may take more than 100 years to reach
the canopy ,and even longer for the making of a closed
stand.
　Since the behavior of wind varies among slopes ,the
prevailing wind is west2south2west [30 ] ,and the distur2
bance pattern is therefore considered different in rela2
tion with topography and slope aspect . It was found
that with this image resolution ,gaps smaller than 100
m in diameter were difficult to discriminate f rom
noise ,hence the image was filtered with 32by23 pixel
window to smooth such pixels. The number of pixels
of coniferous forest zone (1 100～1 700 m asl) was 1
215 622 (about 109 000 hm2) . The spatial variation
that there were tremendous patches of birch forest
was the characteristics of the vegetation on the west2
ern slope. This is considered the result of wind effects
because of the difference in wind resistance among
species[10 ]. As an example , in 1987 , we investigated
the windthrow area formed in 1986. In the birch2
mixed coniferous forest , Picea and A bies were almost
entirely thrown down , while there were numerous
stems of mountain birch remained standing. As a
whole ,the change of vegetation in the two directions
were in a balance ,showing that the forest community
as a continuum was in a stable status.
4 　DISCUSSION
　Picea and A bies presented a pattern of continuous
regeneration. There were a large number of young in2
dividuals ,which were illust rated in both DBH dist ri2
bution ( Fig. 1 and Fig. 2) and age structure[8 ] . A bies
is a typical shade demanding species. The seedlings
under closed canopy can reach 2 000 stems·hm - 2 ,
and the mortality ratio is considerably high under the
age of 50 years. Contrary to this ,the age structure of
8111 应 　用 　生 　态 　学 　报 　　　　　　　　　　　　　　　　　　15 卷
larch was characterized by the high proportions in
large sizes and old ages ,most individuals were > 25 m
in height and > 120 years in age. This pattern results
f rom the lack of succeeding layer [17 ] ,which is differ2
ent f rom shade tolerant species.
　Larch is a shade intolerant species as the light satu2
ration of photosynthesis is higher than 60 000 lux.
Contrary to this ,the saturation point of A bies is only
10 000 [29 ] . It is clear that the regeneration of larch
occurs in open land where the high radiation benefits
its growth. On the other hand , the regeneration of
shade tolerant species like Picea and A bies needs
shades made by pioneer species. In other words ,larch
provides suitable conditions for the regeneration of cli2
max species.
　The role of mountain birch is a little different f rom
other species ,which has a st rong adaptability to light
condition. This can be demonstrated by its continuous
DBH dist ribution ( Fig. 3) and the large amount of
saplings and seedlings in the forest . When a gap is
created ,it can invade in the open area as a pioneer ,
and provides proper conditions for the regeneration of
spruce and fir. Different f rom larch ,it can grow under
the coniferous canopy[10 ] . In brief , mountain birch is
not only a pioneer ,but also a coexistent species in the
upper subalpine coniferous forest .
　The coniferous forest was dotted with many patch2
es that were dominated by pioneer species ,like L ari x
olgensis and Bet ula erm anii , the former can appear
at any elevation ,while the latter is mainly dist ributed
above 1 500 m asl. According to our field investiga2
tion , L ari x will be replaced by the climax species of
Picea and A bies . This is the case in a specific plot or
patch when the pioneer phase is gradually turning to
the climax stage. On the other hand ,the so2called cli2
max parts are vulnerable to wind disturbance. The ev2
ergreen species have a large amount of leaf
biomass[9 ] , which weakens its resistance to wind
blowing. In other words , when the patches in sec2
ondary phase are replaced by climax phase , natural
disturbance like wind throw would create new gaps.
Thus the mosaic st ructure of the coniferous forest
zone is maintained.
　It is believed that wind disturbance in different
scales would repeatedly happen during the restoration
of vegetation ,although its f requency at each scale re2
quires further studies. In the coniferous forest zone ,
gaps are always created , as demonstrated by the re2
mote sensing approach ( Fig. 5) . This is in a balance
with progressive succession. In other word , the two
sucessional directions offset each other. It is then con2
cluded that the climax of the subalpine is in fact a mo2
saic st ructure , consisting of patches in different
stages. It is evident that the mosaic st ructure is main2
tained by disturbance ,and the gap formation is coun2
teracted by patch building. The coniferous forest is
dominated by climax species and is always disturbed
by wind throw. After disturbance ,the pioneer species
invade the site ,and the forest develops toward the cli2
max.
　The phrase“climax”is defined as a relatively stable
ecological stage that is achieved through successional
adjustment to an environment ,or the final stage in e2
cological succession. This concept is often opposed ,be2
cause such a“stable community”consisting of various
plant species is inexistent . When we predict the sta2
bility or dynamics of a plant community , the conclu2
sion largely depends on the scale of the community we
refer to . It is hard to consider that the subalpine will
be covered by vegetation other than the current conif2
erous forest . A permanent plot may show us dramatic
variability in species composition and size st ructure
with time ,but if we take the whole community as a
continuum of plant association ,it is stable or in a sta2
tus of equilibrium.
REFERENCES
1 　Chavez PSJ and Mackinnon DJ . 1994. Automatic detection of vege2
tation changes in the southwestern United States using remotely
sensed images. Photo Eng Rem Sen ,60 :571～583
2 　Chen L ,Bao X and Li C. 1964. Some structural characteristics of
the main plant communities in the vertical zones of Changbai moun2
tain. Acta Phyto Geobot S in ,2 :207～225 (in Chinese)
3 　Chen X ,Li BL and Lin ZS. 2003. The acceleration of succession for
the restoration of the mixed2broadleaved Korean pine forests in
Northeast China. For Ecol M an ,177 :503～514
4 　Chertov O , Komarov A , Kolstrom M , et al . 2003. Modelling the
long2term dynamics of populations and communities of trees in bo2
real forests based on competition for light and nitrogen. For Ecol
M an ,176 :355～369
5 　Howarth PJ and Wickware GM. 1981. Procedures for change de2
tection using Lansat digital data. Int J Rem Sen ,2 :277～291
6 　Jarvis C. 1994. Modeling forest ecosystem dynamics using multi2
temporal multispectral scanner (MSS) data. A dv S pace Res ,14 :277
91117 期 　　　　　　　　　　　刘琪　等 :利用影像判读与群落监测分析长白山针叶林动态 　　　　　　　　　　　
～281
7 　Jensen J R and Toll DL . 1982. Detecting residential land2use devel2
opment at the urban fringe. Photo Eng Rem Sen ,48 :629～643
8 　Li W ,Deng K ,Shao B , et al . 1995. Studies on the community char2
acteristics of dark coniferous forest of Changbai mountain. Res For
Ecol ,7 :1～15 (in Chinese)
9 　Li W ,Deng K and Li F. 1981. Study on biomass and primary pro2
duction of main ecosystems in Changbai mountain. Res For Ecol ,2 :
34～50 (in Chinese)
10 　Liu QJ . 1997. Structure and dynamics of the subalpine coniferous
forest on Changbai mountain ,China. Plant Ecol ,132 :97～105
11 　Liu QJ ,Dai L and Chen H. 1998. Changes of community character2
istics of a broad2leaved2conifer mixed forest after selection. J For
Res ,9 :152～159
12 　Miller AB ,Bryant ES and Birnie RW. 1998. An analysis of land
cover changes in the Northern Forest of New England using multi2
temporal Landsat MSS data. Int J Rem Sen ,19 :245～265
13 　Liu QJ and Takeuchi N. 2001. Vegetation inventory of a temperate
biosphere reserve in China by image fusion of Landsat TM and
SPOT HRV. J For Res ,6 :139～146
14 　Miyadokoro T , Nishimura N and Yamamoto S. 2003. Population
structure and spatial patterns of major trees in a subalpine old2
growth coniferous forest ,central Japan. Fort Ecol M an ,182 :259～
272
15 　Mouat DA ,Mahin GG and Landcaster J . 1993. Remote sensing tech2
niques in the analysis of change detection. Geocarto Int , (2) :39～49
16 　Muchoney DM and Haack BN. 1994. Change detection for monitor2
ing forest defoliation. Photo Eng Rem Sen ,60 :1243～1251
17 　Okitsu S ,Ito K and Li C. 1995. Establishment processes and regen2
eration patterns of montane virgin coniferous forest in northeastern
China. J Veg Sci ,6 :305～308
18 　Palmer AR and van Rooyen AF. 1998. Detecting vegetation change
in the southern Kalahari using Landsat TM data. J A rid Envi ron ,
39 :143～153
19 　Primack RB , Ashton PS , Chai P , et al . 1985. Growth rates and
population structure of Moraceae trees in Sarawak ,east Malaysia.
Ecology ,66 :577～588
20 　Pulido FJ ,Diaz M and de Trucios SJ H. 2001. Size structure and re2
generation of Spanish holm oak Quercus ilex forests and dehesas :
effects of agroforestry use on their long2term sustainability. For
Ecol M an ,146 :1～13
21 　Quarmby NA and Cushnie JL . 1989. Monitoring urban land cover
changes at the urban fringe from SPOT HRV imagery south2east
England. Int J Rem Sen ,10 :953～963
22 　Saxena A K and Singh J S. 1984. Tree population structure of certain
Himalayan forest associations and implications concerning their fu2
ture composition. Vegetatio ,58 :61～69
23 　Singh A. 1989. Digital change detection techniques using remotely2
sensed data. Int J Rem Sen ,10 :989～1003
24 　Sunar F. 1998. An analysis of changes in a multi2date data set : a
case study in the Ikitelli area , Istanbul , Turky. Int J Rem Sen ,19 :
225～235
25 　Vandermeer J H ,Boucher DH ,de la Cerda IG , et al . 2001. Growth
and development of the thinning canopy in a post2hurricane tropical
rain forest in Nicaragua. For Ecol M an ,148 :221～242
26 　Wang J R ,Letchford T ,Comeau P , et al . 2000. Above2 and below2
ground biomass and nutrient distribution of a paper birch and sub2
alpine fir mixed2species stand in the Sub2Boreal Spruce zone of
British Columbia. For Ecol M an ,130 :17～26
27 　Wang Z , Xu Z ,Li X , et al . 1980. The structure of major forest
types in northern slope of Changbai Mountain. Res For Ecol ,1 :25
～42 (in Chinese)
28 　Yang H and Xie H. 1994. Study on the reconstruction of distur2
bance history of Pinus koraiensis mixed forest in Changbai moun2
tain. Acta Phyt S in ,18 :210～208 (in Chinese)
29 　Yao J ,Lin J and Yin Z. 1983. The biological characteristics of Pi2
nus koraiensis with special reference to its adaptability to light . Res
For Eco ,3 :258～264 (in Chinese)
30 　Zhang F ,Chi Z and Li X. 1980. Preliminary analysis on the climate
of Changbai mountain area. Res For Ecol ,1 :193～213 (in Chinese)
31 　Zhao D. 1980. Vertical zonation of Mt . Changbai. Res For Ecol ,1 :
65～70 (in Chinese)
Author introduction 　L IU Qijing ,male ,doctor degree ,profes2
sor. Major at forest ecology and vegetation remote sensing. Tel :
0102648829027 ,E2mail :liuqijing @hotmail. com.
0211 应 　用 　生 　态 　学 　报 　　　　　　　　　　　　　　　　　　15 卷