基于度量误差方法的油松林分生长模型

Stand growth model for Pinus tabulaeformis using measurement error method

为掌握北京地区油松Pinus tabulaeformis生长过程,建立相容的断面积和蓄积模型,通过引入间伐林分与未间伐林分的哑变量,分别建立林分断面积、林分蓄积生长模型,然后从相容的角度出发,建立林分断面积、蓄积量的误差变量联立方程组,并与不含哑变量的传统生长模型的误差变量联立方程组进行比较。经检验,传统误差变量联立方程组中林分断面积和林分蓄积生长模型的预测精度都在92%以上,对油松林分断面积的预测精度高达0.921 5,决定系数高达0.900 1,对油松林分蓄积量的预测精度达到了0.928 3,决定系数高达0.912 3,而引入哑变量的误差变量联立方程组中,模型的预测精度和确定系数稍高,均在93%以上,对油松林分断面积的预测精度高达0.939 8,决定系数达到了0.927 9,对油松林分蓄积量的预测精度在0.930 0以上,决定系数达0.932 8。这说明引入哑变量,一定程度上提高了模型的预测精度,而且所建模型比较合理,形式相对简单,便于应用,不仅使得林分水平上的林分断面积、蓄积量的预测结果具有相容性,同时还考虑了间伐措施对林分生长的影响,达到了林分生长与收获模型整体化研究的目的,为林分的经营管理提供了可靠依据。

To grasp the growth process of Pinus tabulaeformis in Beijing and to establish general and compatible models, stand basal area and volume growth models with dummy variables of thinned and unthinned stands were established. Then from the perspective of compatibility, simultaneous equations with error variables were established for stand basal area and volume. These were compared to simultaneous equations with error variables of the traditional growth model with no dummy variable. Results after inspection showed prediction accuracies of stand basal area and volume were over 92% in the traditional simultaneous equations with error variables. Prediction accuracies of P. tabulaeformis for stand basal area reached 0.921 5 with coefficients of determination as high as 0.900 1; whereas, prediction accuracies of stand volume reached 0.928 3 with coefficients of determination of 0.912 3. However, in simultaneous equations with error variables of introduced dummy variables, the prediction accuracies and coefficients of determination of models were slightly higher with over 93%.Prediction accuracy of stand basal area was over 0.939 8 with coefficients of determination reaching 0.927 9.Prediction accuracy of stand volume was over 0.930 0 with coefficients of determination as high as 0.932 8.Thus, to improve prediction accuracies, models using dummy variables were reasonable as the form was relatively simple, easy to apply, made forecast results of stand basal area and volume predictions consistent, considered the effect of thinning measures on stand growth, and achieved the purpose of stand growth and yield model integrative research, thereby offering a reliable basis for forest management.