PREDICTIONG OF EUCALYPTUS WOOD BY COKRIGING, KRIGING AND REGRESSION

In the Gypsum Pole of Araripe, semiarid zone of Pernambuco, where is produces 97% of the plaster consumed in Brazil, a forest experiment with 1875 eucalyptus was cut off and all the trees were rigorously cubed by the Smalian method. The location of each tree was marked on a Cartesian plane, and a sample of 200 trees was removed by entirely random process. In the 200 sample trees, three estimation methods for variable volume timber, regression analysis, kriging and cokriging were used. To cokriging method, the secondary variable was the DBH (Diameter at Breast Height), and for the regression model of Spurr or the combined variable, it uses two explanatory variables: total height of the tree (H) and the DBH. The variables volume and DBH showed spatial dependency. To compare de methods it was used the coefficient of determination (R2) and the residual distribution of the errors (real x estimated data). The best results were achieved with the Spurr equation R2= 0.82 and total volume estimated 166.25 m3. The cokriging provided and R2 = 0.72 with total volume estimated of 164.14 m3 and kriging had R2 = 0.32 and the total volume estimated of 163.21 m3. The real volume of the experiment was 166.14 m3.

位于巴西伯南布哥州半干旱区的阿拉里皮石膏产区（该产区供应巴西97%的消费用石膏），开展了一项针对1875株桉树的森林试验：试验中所有林木均被砍伐，并通过斯迈尔法（Smalian method）进行严格的材积切块处理。每株林木的位置均被标记在笛卡尔平面坐标系上，并通过完全随机抽样法抽取200株林木作为样本。针对这200株样本林木，采用三种方法开展立木材积估算，分别为回归分析、克里金（kriging）法及协同克里金（cokriging）法。其中，协同克里金法的辅助变量为胸径（DBH, Diameter at Breast Height）；而斯普尔（Spurr）回归模型及组合变量模型则采用两个解释变量：林木总高（H）与胸径。材积与胸径均表现出空间依赖性。为对比三种方法的性能，本研究采用决定系数（R²）以及误差残差分布（实测值与估算值对比）作为评价指标。结果显示，斯普尔回归方程取得最优效果，其决定系数R²为0.82，估算总材积为166.25立方米；协同克里金法的决定系数R²为0.72，估算总材积为164.14立方米；克里金法的决定系数R²为0.32，估算总材积为163.21立方米。该试验的实际总材积为166.14立方米。