应用地径对小兴安岭主要针叶树种胸径和材积预测模型的构建  

Constructing Prediction Models for Diameter at Breast Height and Volume of Major Conifer Species in the Xiaoxing’an Mountains by Applying Stump Diameter

作　　者：孙行丽 张兹鹏 姜立春[1] Sun Hangli;Zhang Zipeng;Jiang Lichun(Northeast Forestry University,Harbin 150040,China)

机构地区：[1]东北林业大学,哈尔滨150040

出　　处：《东北林业大学学报》2023年第12期60-65,73,共7页Journal of Northeast Forestry University

基　　金：国家自然科学基金项目(32271866)。

摘　　要：以黑龙江省伊春市五营林业局丽林实验林场的小兴安岭3个主要针叶树种(落叶松(Larix gmelinii)、红皮云杉(Picea koraiensis)、臭冷杉(Abies nephrolepis))为研究对象,共选取472株解析木数据(树种、胸径、树高、地径等基本信息),采用平均断面积区分求积法计算解析木材积;选取线性、二项式、幂函数3种方程形式构建胸径-地径模型,选取线性、二项式、幂函数、指数函数、自然对数的方程形式构建材积-地径模型,采用加权回归法对模型异方差进行校正;选择平均偏差、平均绝对误差、预测平均百分比误差、平均绝对百分比误差4个指标,采用留一交叉法对所有模型进行预测精度评价;引入虚拟影响因素,采用非线性方差检测方法,检验不同树种间胸径-地径关系的差异、材积-地径关系的差异。结果表明:①分别采用线性、二项式、幂函数,拟合了3个树种的胸径-地径间的相关关系,其中二项式模型的预测精度最高。②分别选取线性、二项式、幂函数、指数函数、自然对数,拟合了3个树种的材积-地径间的相关关系,其中二项式模型的预测效果最好。③选择材积预测值作为误差方差函数评价指标,幂函数、指数函数、常数加幂函数分别作为臭冷杉、落叶松、红皮云杉的误差方差函数,对材积-地径模型(二项式)异方差校正的效果最好。④F检验结果表明,臭冷杉与落叶松之间、落叶松与红皮云杉之间,预测模型存在极显著差异,需要单独构建胸径-地径模型、材积-地径模型;臭冷杉与红皮云杉之间,预测模型不存在差异,可共用一套参数估计值。⑤综合评价结果,二项式胸径-地径模型、二项式材积-地径模型,能够较好反映3个树种的胸径与地径的关系、材积与地径的关系,都适合于研究地区利用地径估算材积;具体应用时,可根据数据特征和研究目的选择合适的方法。A total of 472 resolved wood data (species, diameter at breast height, tree height, stump diameter and other basicinformation) were selected from three major coniferous species ( Larix gmelinii, Picea koraiensis and Abiesnephrolepis ) in the Xiaoxing’an Mountains of Lilin Experimental Forest, Wuying Forestry Bureau, Yichun City,Heilongjiang Province, to calculate resolved wood product using the average cross-sectional area differentiatedproduct method. The average cross-sectional area differentiated product method was used to calculate the volume ofresolved wood;three equation forms of linear, binomial, and power function were selected to construct the breastdiameter-stump diameter model, and three equation forms of linear, binomial, power function, exponential function,and natural logarithm were selected to construct the volume-stump diameter model, and the weighted regressionmethod was used to correct for the heteroscedasticity of the model;and four types of model were selected: the meanbias, the mean absolute error, the mean predicted average percentage error, the mean absolute percentage error, andthe mean predicted average percentage error. The mean bias, mean absolute error, mean absolute percentage error,mean percentage error of prediction, and mean absolute percentage error of prediction were selected to evaluate theprediction accuracy of all the models using the leave-one-out method. Dummy variables were introduced, andnonlinear variance detection methods were used to test the differences in the relationship of diameter at breast heightand volume of wood and stump diameter among different tree species. The results showed: (1) Linear, binomial, andpower function, respectively, were chosen to fit the correlation between breast diameter-stump diameter of three treespecies, among which the binomial model had the highest prediction accuracy. (2) Linear, binomial, power function,exponential function, and natural logarithm were selected to fit the correlation between wood volume and stumpdiameter of three tre

关 键 词：针叶树种 胸径-地径模型 材积-地径模型 虚拟影响因素 

分 类 号：S758.1[农业科学—森林经理学] S758.5[农业科学—林学]