CAJ | 학술논문

以马尾松人工林132株优势木树干解析数据为训练样本，用145块标准地优势木平均高数据为检验样本，把林分年龄和地位指数或优势木平均高作为输入变量，将优势木平均高或地位指数作为输出变量，通过构建人工神经网络逆模型的途径，分别建立了多形地位指数曲线式和计算式模型。结果表明，多形地位指数曲线式的总体拟合精度为99.64％，总体预测精度达96％以上，比传统技术构建的多形地位指数模型能较真实地模拟各地位级的多形曲线；多形地位指数计算式的总体拟合精度为98.81％，用于计算地位指数，省去用迭代法计算地位指数的工作量。基于BP神经网络模型多形地位指数模型，对马尾松人工林地位指数测定提供指导作用，可为森林立地质量评价提供理论依据。
이마미송인공림132주우세목수간해석수거위훈련양본，용145괴표준지우세목평균고수거위검험양본，파림분년령화지위지수혹우세목평균고작위수입변량，장우세목평균고혹지위지수작위수출변량，통과구건인공신경망락역모형적도경，분별건립료다형지위지수곡선식화계산식모형。결과표명，다형지위지수곡선식적총체의합정도위99.64％，총체예측정도체96％이상，비전통기술구건적다형지위지수모형능교진실지모의각지위급적다형곡선；다형지위지수계산식적총체의합정도위98.81％，용우계산지위지수，성거용질대법계산지위지수적공작량。기우BP신경망락모형다형지위지수모형，대마미송인공임지위지수측정제공지도작용，가위삼림입지질량평개제공이론의거。
By means of constructing inverse model of artificial neural network , the polymorphic site index curve and calculation formula were established with stem analysis data of 132 dominant trees of Pinus massoniana planta-tion as training samples , and with average height of 145 dominant trees from standard field as tested samples.The stand age and site index or average high of dominant tree was selected as input variables , and the average height of dominant tree or site index was regarded as output variables , The results showed that the overall accuracy of the in-dex curve was 99.64 %, the overall prediction accuracy reached 96 %, and it could precisely simulate polymor-phic curve than the polymorphic site index model constructed through traditional technology.The overall accuracy of the calculation formula was 98.81 %, and it could reduce workload to calculate position index by using iterative method.