【摘要】 以内蒙古地区牙克石林业管理局的大兴安岭天然白桦林（Betula platyphylla）为研究对象，利用198块样地数据分析天然白桦林林分碳密度与各林分变量之间的关系，建立碳密度预测基础模型，同时将立地条件（草类白桦林、杜鹃-越桔白桦林、榛子白桦林）作为哑变量引入到预测模型中，对不同林型的林分碳密度进行预测，为林业研究中碳密度模型的构建以及森林碳汇工作提供思路和方法。结果表明，天然白桦林林分碳密度基础模型决定系数（R~2）为0.703，均方根误差（RMSE）为8.615 t/hm~2，赤池信息量（Akaike information criterion,AIC）为841.206，贝叶斯信息量（Bayesian Information Criterion,BIC）为851.071。引入立地条件哑变量后，R~2有所增大，最大达到0.818,RMSE均小于等于8.241 t/hm~2，说明模型具有较好的稳定性，预估参数较为精确。哑变量模型的AIC均小于等于541.431,BIC均小于等于550.320。哑变量模型能够反映不同立地条件下碳密度的变化，在模型的拟合和检验方面都显示适合于研究地区林分碳密度的预测，为天然白桦林碳密度估算提供参考。更多还原

【Abstract】 This study focused on the natural birch forest（Betula platyphylla） in the Daxing’an Mountains of Yakeshi Forestry Administration in Inner Mongolia. Using 198 sample plots, the relationship between carbon density in the natural birch forest stand and various stand variables was analyzed. A basic model for predicting carbon density was established, incorporating site conditions（grass birch forest, rhododendron-vaccinium birch forest, hazelnut birch forest） as dummy variables to predict stand carbon density across different forest types. This paper offered insights and methodologies for constructing carbon density models and advancing forest carbon sequestration in forestry research. The results showed that the determination coefficient（R²） for the basic model of carbon density in the natural birch forest was 0. 703, root mean square error(R_MSE) was 8. 615 t/hm^-2, Akaike information criterion（AIC） was 841. 206, and Bayesian Information Criterion（BIC） was 851. 071. After site conditions were introduced as dummy variables, R² increased to a maximum of 0. 818, and RMSE were all less than or equal to 8. 241 t/hm^-2, indicating that the model had good stability and the predicted parameters were more accurate. AIC for the dummy variable model was less than or equal to 541. 431, and BIC was less than or equal to 550. 320. The dummy variable model can reflect the change of carbon density under different site conditions, and both the fitting and testing of the model show that it is suitable for the prediction of forest carbon density in the study area, which provides a reference for the estimation of natural birch forest carbon density.更多还原