應用不同結構樹高曲線式模擬臺灣杉人工林之效果評估

Assessing Prediction Effects among Height-Diameter Models with Varied Structures for a Taiwania (Taiwania cryptomerioides Hayata) Plantation

林政融(Zheng-Rong Lin) ； 顏添明(Tian-Ming Yen)

《台灣林業科學》 36卷2期 (2021/06) Pp. 111-125

https://doi.org/10.7075/TJFS.202106_36(2).0002

摘要

樹高曲線式（height-diameter（H-D）model）係採用胸高直徑（diameter at breast height（DBH））推估樹高（tree height（H））的重要工具，然而H之模擬效果會隨著H-D model的結構而改變。本研究旨在探討模式結構對H-D model模擬表現之影響。研究區域位於臺灣中部地區惠蓀林場之臺灣杉（Taiwania cryptomerioides Hayata）人工林林分，共獲104株具DBH與H之單木資料。本研究採用不同種模式型態之H-D model進行建模，採用residual sum of squares（RSS）、root mean square error（RMSE）、Akaike information criterion（AIC）及relative rank（R-rank）等指標評估模式。並以成對樣本t-test（paired sample t-test）及二因子變異數分析（two-way analysis of variance（ANOVA））分析模式模擬之效果。結果顯示，在所有模式中H = a+ bD + cD^2 + d log D表現最佳。而非線性模式方面，約束模式通過原點可提升模擬效果；然而在線性模式方面，3及4參數模式模擬結果較2參數為佳。比較2種模式型態在參數間的模擬效果，非線性模式在2參數結果較佳，而在3及4參數則與線性模式效果相同。

關鍵字

約束模式 ； 惠蓀林場 ； 參數數目 ； 模式型態 ； 臺灣杉

並列摘要

The height-diameter (H-D) model is an important tool for predicting tree height (H) based on the diameter at breast height (DBH). However, the performance of the H-D model varies with the model structure. The purpose of this study was to examine the performances of H-D models with various model structures. The research site was located in central Taiwan. Data were collected from a Taiwania (Taiwania cryptomerioides Hayata) plantation at the Huisun Forest Station, and in total, the DBH and H of 104 individual trees were obtained. We adopted various H-D models with different structures to establish the models. The residual sum of squares (RSS), root mean square error (RMSE), Akaike information criterion (AIC), and relative ranking (R-rank) performance criteria were employed as criteria. A paired sample t-test and two-way analysis of variance (ANOVA) were used to assess model performances. Results showed that H = a + bD + cD^2 + d log D stood out among all models. Nonlinear models had better performances when they were constrained to pass through the origin. In linear models, the performances of 3- and 4-parameter models were better than those of 2-parameter models. In a comparison of the number of parameters between models, nonlinear models performed better than linear models at the 2-parameter level due to large biases in the linear models.

並列關鍵字

constrained model ； Huisun Forest Station ； number of parameters ； model type ； Taiwania