利用指標克利金法和一般克利金法建立樹種的空間分佈

Application of Indicator Kriging and Ordinary Kriging to Map Spatial Distribution of Trees Species

黃仁梅(Ren-Mei Ooi)

指導教授 ： 林增毅

共同指導教授 ： 澤大衛(David Zelený)

國立臺灣大學/生物資源暨農學院/森林環境暨資源學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201803037

摘要

了解森林的樹種組成、林木大小和樹種的空間分佈對森林經營者而言，是非常重要的。森林經營者可以利用林分產生器來建立林分結構資料，並將這些資料用在森林的經營規劃上。目前林分產生器在樹種分配上大多是採取隨機分配的方式，但是樹種在森林裡的分佈不完全是隨機的，而是會受到其他因素的影響。因此林分產生器的樹種分配方式有需要改善的部分。Hershey 等人在1997年的研究指出指標克利金法適合用於模擬樹種在森林裡的空間分佈。因此，在本次研究中，我們將使用指標克利金法和一般克利金法來模擬福山森林動態樣區其中三種樹種各自的空間分佈。指標克利金法所預測的是樹種出現的機率，而一般克利金法則是預測樹種的數量。為了了解取樣強度、樣區大小和取樣方法對指標克利金法及一般克利金法模擬的影響，本次研究也以四種取樣強度（分別為5 %, 10 %, 20 %, 50 %）、兩種樣區大小（5×5 m 和 10×10 m) 和兩種取樣方法（簡單隨機取樣和系統取樣）組成16個組合來進行兩種克利金法的模擬。研究結果表示指標克利金法和一般克利金法可以準確預測本研究所選擇三種樹種的空間分佈。取樣強度對兩種克利金法的模擬有明顯的影響。當取樣強度越強，模擬的精準度就會越高。而樣區大小和取樣方法對模擬的影響並不是非常明顯且需要更多的探討。未來森林經營者可以將這兩種克利金法和林分產生器作結合，以便建立更趨近於真實林分的林分結構資料。

關鍵字

樹種空間分佈 ； 半變異圖 ； 克利金法 ； 林分產生器 ； 點過程 ； 地理空間統計

並列摘要

Understanding species composition, size, and spatial distribution of trees are important for forest managers. Recent studies state that forest managers can use stand generators to generate stand structure information. However, method of species assignment of stand generators must be improved. Hershey et al. (1997) introduced indicator kriging to predict spatial distribution of species and suggested this method was a suitable tool for the tree species spatial distribution prediction. In this study, we used indicator kriging to predict spatial presence/absence and ordinary kriging to predict abundance of three tree species in Fushan Forest Dynamics Plot. To study influence of sampling intensity, cell size, and sampling method on prediction accuracy, four levels of sampling intensities (i.e. 5 %, 10 %, 20 %, and 0 % of total number of cells), two different cell sizes (i.e. 5×5 m and 10×10 m), and two different sampling methods (i.e. simple random sampling and systematic sampling) were included in the kriging estimation. Thus, there were a total of 16 combinations of cell size, sampling method and sampling intensity for indicator kriging and for ordinary kriging. Result of this study showed that indicator kriging and ordinary kriging were well behaved in predicting spatial distribution of the three tree species case studies. Sampling intensity significantly influenced the kriging prediction accuracy, by increasing the prediction accuracy as it increases. The influence of cell sizes and sampling methods on indicator kriging and ordinary kriging were smaller than sampling intensity and require further investigation. Forest managers can use these interpolation methods to simulate forest stand for decision making in forest management.

並列關鍵字

Species spatial distribution ； Semivariogram ； Kriging ； Stand generator ； Point process ； Geospatial statistics

參考文獻

Andersen, K.M., Turner, B.L., Dalling, J.W., 2010. Soil‐based habitat partitioning in understorey palms in lower montane tropical forests. Journal of Biogeography 37, 278-292.

Arslan, H., 2012. Spatial and temporal mapping of groundwater salinity using ordinary kriging and indicator kriging: The case of Bafra Plain, Turkey. Agricultural Water Management 113, 57-63.

Barreto, P., Amaral, P., Vidal, E., Uhl, C., 1998. Costs and benefits of forest management for timber production in eastern Amazonia. Forest Ecology and Management 108, 9-26.

Biging, G., Robards, T., Turnblom, E., Van Deusen, P., 1994. The predictive models and procedures used in the forest stand generator (STAG). Hilgardia 61, 1-36.

Bivand, R., Pebesma, E., Gómez Rubio, V., 2013. Applied spatial data analysis with R. Springer New York, New York.