臺灣杉經疏伐人工林之林分直徑分佈模型與預測

Abstract

直徑分佈模型可用來了解當下及未來可能的森林林分蓄積量，以幫助管理人擬定育林計劃，而機率密度函數（probability density functions，pdf）為最常用來擬合（fit）和模擬（model）直徑分佈模型趨勢的方法。本研究使用兩個參數的Weibull和Johnson’s SB分佈的機率密度函數模擬六龜地區經疏伐臺灣杉人工林的直徑分佈模型，包含估算擬合直徑分佈模型時兩分佈的參數、使用線性迴歸模型（linear regression model）分析參數與已知林分性態值（stand characteristic）的關係和使用赤池信息量（Akaike’s information criterion，AIC）、AICc和貝葉斯信息量（Bayesian information criterion，BIC）比較兩分佈的擬合優良性（goodness-of-fit）。 結果顯示兩種分佈的形狀參數值皆隨林齡增加些微下降，尺度參數值則皆隨林齡增加大幅上升；由各參數離勢（variation）變化可知直徑分佈在疏伐處理後原本並不穩定，但隨時間漸漸趨於穩定。我們也發現多種林分性態值與機率密度函數中的參數呈線性關係，例如：是否經疏伐處理、疏伐強度、疏伐後年數、 疏伐前林分密度和疏伐前後林分斷面積等等；AIC、AICc和BIC結果則顯示 Johnson’s SB 分佈在擬合時有比較好的結果，表示Johnson’s SB 分佈在擬合我們的資料時比兩個參數的Weibull分佈彈性。

Diameter distribution can be used to understand the total growing stock of a forest stand at present and possibly in the future, assisting managers to plan silvicultural treatments. Probability density functions (pdf) are commonly used to fit and model trends of diameter distribution. In this study, parameters in two-parameter Weibull and Johnson’s SB probability density function were estimated and compared when fitted to the diameter distributions of thinned and non-thinned taiwania plantations in Liukuei, and the linear regression models were used to predict the relationships between parameters and stand characteristics. The results showed that shape parameters decreased slightly with stand age while scale parameters increased greatly. The large variations of all parameters indicate that the diameter distribution is initially unstable following thinning treatments but tend to stabilize over time. We further found some stand characteristics such as thinning treatments, thinning intensity, year after thinning treatments, stand density and basal area were related to the parameters in the probability density function. The comparison results in AIC, AICc and BIC indicated Johnson’s SB as more flexible when fitting our diameter data of 35 years old plantations compared to 2Weibull.