Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/65559
標題: 森林地景監測之研究─以惠蓀林場為例
Study on Landscape Monitoring of Forest-An Illustration of Huisun Experimental Forest Station
作者: 黃志成
Huang, Chih-Cheng
關鍵字: Forest ecosystem;森林生態系;Forest resource monitoring;Fractal theory;森林資源監測;碎形理論
出版社: 森林學系
摘要: 
森林生態系經營理念之興起,引導林業經營朝向永續性及生態性的方向發
展。不管行政部門或研究單位,都將以蒐集完整而長期的森林生態系資料
為導向,透過資料管理和分析以符合生態系經營的原則。監測系統的建立
與研究是達到森林合理經營的必要途徑,本文以監測系統的定義為出發點
,以狀態、變遷及預測等三個步驟為研究主題,導入地景生態學的觀念,
首先以土地利用地景單元建立資料庫,進行變遷分析,並透過機率模式進
行地景預測。其次以空間模式為工具,使用具有區域變數理論為基礎之
Kriging,推估林分結構母數,再利用來自地景生態學之形狀指數為區?
(一)藉由相關研究報告之蒐集與歸納,研擬土地利用地景監測系統之
理論架構,包括資料蒐集子系統、資料庫子系統、分析模式子系統及決策
支援子系統。
(二)以惠蓀林場1964、1977及1987年土地利用型圖之製作為實例,探
討圖籍數化之相關問題,供作地景製圖上之參考。而透過GPS對永久樣區
林木位置及造林台帳資料之差分定位測量,由圖形套疊結果及DGPS精度來
看,以GPS及GIS結合運用,可做為土地利用型圖(地景鑲嵌圖)製作之輔助
工具。
(三)利用GIS建立空間位相關係可以迅速查詢地景單元之相關資訊,
建立不同年代之地景結構資料庫,並以地景生態學相關假說而建立面積、
周長、頻度、百分比、形狀指數、內緣比等地景基本資料。經由不同年代
變遷分析的結果可知惠蓀林場三個年代之平均周長及頻度均逐年上升;形
狀指數及內緣比亦逐年增加,表示出林場地景單元發展潛勢及邊緣效應增
加的趨勢。此一結果,可協助森林經營者瞭解林場地景單元之變化情形。
(四)在地景生態指標的變化上,本研究求算地景單元優勢度、覆蓋度
、歧異度、最大歧異度及均勻度指數等,分析地景的變化作為地景管理之
參考。並透過統計檢定,偵測歧異度指數之差異性,其中1964-1977年間
具顯著差異,而1977-1987年間則無顯著差異,顯示地景異質性變化由第
一期的13年間明顯上升,而至第二期的十年間逐漸趨緩。
(五)地景單元透過轉移矩陣之建立及運算,可發現檜木、杉木、闊葉
樹林型、裸露地和建地具有較高之周轉率,若能描繪地景單元變化路徑圖
,將更有助於地景變化之瞭解與控制。
(六)以Markov機率模式建立地景的預測模式,透過第一、二期資
料(1964、1977年)驗證第三期資料(1987年),可知土地利用型(地景單元)
之面積平均誤差為54.97公頃(約0.74%),誤差來源係因造林地等人工引入
嵌塊體之不確定因素造成,顯示Markov機率模式在預測較頻繁之地景變化
時,仍有待改善以儘可能減少誤差。
(七)以第三次資源調查之大安溪事業區地面樣區資料進行先期
Weibull機率密度函數之母數推估。再透過Kriging空間推估模式進行點推
面之程序,所建立之空間結構母數可於GIS上繪製等值線,進而描述全區
之林分結構分布情形,透過本方法之介紹及演算可供林業界空間資料分析
之參考。
(八)區域化變數理論可透過半變異之求配繪製變異元圖,描述區域化
變數之空間變異結構。而碎形理論可以解釋空間中不規則的現象,尤其是
布朗碎形函數可以描述空間中區域變數之無規律分布現象。本研究以地景
單元之結構特性及其形狀指數,應用前述二項理論完成以下數項之研究:
1.地景單元碎形分析
(1)地景單元之碎形分析中指出惠蓀林場地景單元碎形維度逐年降低
,分別為1964年之1.588、1977年之1.4367及1987年之1.3732;而東北角
海岸風景特定區則逐年增加,即1983年之1.8973至1993年增加為1.9495。
主要原因在於二種地景性質之差異,惠蓀林場之土地利用狀況以森林覆蓋
為主,而東北角海岸風景特定區則包含了複雜的土地利用類型,人為干擾
因素較多,造成整體破碎及複雜程度的提高。
(2)為分析惠蓀林場碎形維度變化之原因,以GIS篩選出林場最為集約
之第三林班,進行碎形維度運算,在三個年代中分別為1964年之1.6751
,1977年之1.6943及1987年之1.7208。碎形維度具有逐年增加之趨勢,驗
證人為干擾將使地景單元之破碎程度增加。經由交叉比對可得知第三林班
之地景與東北角風景特定區之地景具有相同之變化方向。而整體而言,若
以碎形維度之變化為準則,來描述地景破碎程度時,則惠蓀林場並未朝向
破碎地景方向發展。
2.在杉木形狀指數之空間分析中,不同的取樣間距(300公尺,400公
尺及500公尺)其理論指數模式分別為
r(h)300=0.0951[1-exp(-h/4.3860)]
r(h)400=0.1031[1-exp(-h/4.1459)]
r(h)500=0.1016[1-exp(-h/3.7608)]
利用區域變數理論反應短距離內空間變異的特性,可以看出若在
容許的取樣成本下lag=300精度較高,而lag=400與lag=500間則差異不大
,透過不同間距空間變異模式之建立即可瞭解地景監測的尺度性質。
3.整合區域變數理論與碎形理論,研究杉木形狀指數之空間尺度,可
以發現影響範圍內布朗碎形性質,在三種取樣間距內具有自我相似性,而
其碎形維度值約在1.4左右,因此杉木形狀指數在此範圍內具有尺度不變
性。
經由前述之研究成果,可提供森林資源監測系統設立及空間資料分析上之參考

Forest ecosystem management concepts led to the sustainable and
ecological development in forest management. The public
agencies and institutions have been oriented to data collection
from a long-term survey in forest ecosystem. The content of
data management and analysis were asked to match the criteria
of forest ecosystem management. The development and research
of monitoring system is a necessary path for future forest
resource management. This paper deals with the definition of
monitoring system in
The summary of the obtained is as follows:
1.With the process of establishment and purpose of the
related data, this paper divides the landscape monitoring
system into four parts, which are
(1) Data collection
(2) Database management
(3) Data analysis
(4) Decision-making support system
2.GPS was used to collect spatial data with coordination.
GIS was developed to store and analyze diverse spatial data.
In the case study, we integrated GPS and GIS techniques in
forest mapping and monitoring. The results showed that
integrating GIS and GPS supports a good method in forest
mapping.
3.The aerial photographs of different periods (1964, 1977,
1987) in land-use were interpreted, and those land-use maps
were digitized in GIS. We estimated the parameters of
landscape units, which are frequency, perimeter, interior-to-
edge ratio and shape index in different period of time. In the
case study, interior-to-edge ratio and shape index were
increased in different periods, which implies that edge effect
has a potential development in landscape units.
4.The ecological parameters of landscape were estimated,
those parameters are Shannon-Weaver diversity index, minimum
diversity, evenness index, and dominant index is derived. T-
test is used as a tool to check the difference of landscape in
different periods. The T-test of Shannon-Weaver diversity
shows that the period of 1996-1977 is significantly different
in the landscape diversity of land use. There is not much
difference between 1977 and 1987. It means the diversity and
change of land-uses incr
5.Transition matrix could detect the change information of
landscape units. The results showed cypress, Chinese fir,
hardwood, logging and building area have a high rate to of
transition in landscape units.
6. The Markov stochastic model is used to predict the
change of landscape units in the three periods. From the
change model of 1964-1977 to predict those of 1987, we found
the difference between estimated value and observed value are
54.97ha (0.74%). The difference is came from plantation which
human activities made important effect on uncertain change in
landscape.
7. In Kriging model were used to estimate the parameter-B
of Weibull pdf in stand structure with spatial distribution.
Exponential model was used to fit the relationship between
distance and semivariance, which could expatiate stand
structure with a contour.
8.The theory of regionalized variable could detect
properties of spatial variability with semivariance and
semivariograms. Fractals have two important characteristics,
which are the idea of self-similarity and the concept of a
fractional dimension. Fractional Brownian functions could be
better approximated for the spatial variations of high non-
regular. We integrated the two theories and applied it to
shape index - the structure parameters of landscape units in
spatial variations and scales. The res
(1) In the paper, fractal theory is used to study and
analyze the spatial characteristics of landscape of Huisun
Experimental Forest Station and Northeast Coast National Scenic
Area. The fractal dimension D of Huisun Experimental Forest
Station in 1964, 1977 and 1987 are 1.5885 in 1964, 1.4367 in
1977 to 1.3732 in 1987, which is decreasing tendency. But the
D value of Northeast Coast National Scenic Area is from 1.8973
of 1983 to 1.9495 of 1993, which is increasing. Human
disturbance made the land-u
(2) The 3rd compartment is the most intensive area in
Huisun Experimental Forest Station. The fractal dimension D of
3rd compartment in 1964, 1977 and 1987 were computed. The
results showed there is an increasing tendency in fractal
dimension D from 1.6751 of 1964, 1.6743 of 1977 and 1.7208 of
1987. It does coincide with the change direction of Northeast
Coast National Scenic Area. The results also indicated that
the fractal dimension D be suitable to describe the landscape
spatial development of hu
(3) In the spatial analysis of shape index to Chinese fir,
original sample data is resampled at different spatial sampling
intervals (300m, 400m and 500m). The exponential models were
used to describe as follows:
r(h)300=0.0951 [1-exp(-h/4.3860)]
r(h)400=0.1031 [1-exp(-h/4.1459)]
r(h)500=0.1016 [1-exp(-h/3.7608)]
The theory of regionalized variable - shape index could
reflect spatial variation in short range. These models
indicate that lag=300 could get higher precision then the two
others in same cost. There are no difference between lag=400
and lag=500.
(4) Spatial scale has been one of most interesting subjects
in natural resources monitoring. Theory of regionalized
variable and fractal theory were used to study the properties
of spatial scale variability of shape index of Chinese fir.
This results showed that shape index of the three sampling
intervals in 300m, 400m and 500m were got the same D value.
Which shows that scale is invariant in the range.
The method of estimating the landscape status, change and
predictions could support forest resource monitoring and
spatial analysis. In practical applications, Kriging spatial
interpolation model was applied point to area estimate for
Weibull B parameter of stand structure. To integrate the
theory of regionalized variable and fractal theory for
analyzing spatial structure and scale of landscape units. We
hope this paper could contribute to long-term ecological
monitoring and decision-making in forest
URI: http://hdl.handle.net/11455/65559
Appears in Collections:森林學系

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.