| 摘要: |
| 【目的】研究华山松种源对树高生长模型参数的影响，为建立通用的华山松树高生长模型奠定基础。【方法】以湖北省宜昌市大老岭林场华山松种源试验林为研究对象，利用林龄4～10年、24年和32年的树高数据，建立华山松11个种源的基础模型、哑变量模型和非线性混合模型，并对各模型的拟合精度进行比较。【结果】在Richards、Logistic、Schumacher和Korf 4个最常用的树高理论方程中，以Richards方程的拟合效果最好，故选其作为基础模型。哑变量模型和非线性混合模型的检验结果均显示，华山松种源对树高生长模型的渐进参数有显著影响，而对生长速率参数和形状参数无显著影响。对比3种模型的决定系数R2、残差平方和SSE、平均绝对残差|E|和均方根误差RMSE 4个精度评价指标发现，哑变量模型和非线性混合模型的精度比基础模型明显提高，二者平均精度的决定系数R2提高1.37%，残差平方和SSE降低18.27%，平均绝对残差|E| 降低12.77%，均方根误差RMSE降低 9.6%；哑变量模型的拟合精度比非线性混合模型稍高，决定系数R2提高0.1%，残差平方和SSE、平均绝对残差|E|、均方根误差RMSE分别降低1.8%，4%和0.9%。【结论】华山松种源对树高生长模型的渐进参数有显著影响，而对生长速率参数和形状参数无显著影响。在种源数量不多时，利用哑变量模型建立通用的华山松种源树高生长模型是一种较好的方法。 |
| 关键词: 华山松 哑变量模型 非线性混合模型 种源 树高生长模型 |
| DOI： |
| 分类号: |
| 基金项目:“十二五”国家科技支撑计划项目（2012BAD01B01） |
|
| Effects of Pinus armandi Franch provenances on parameters of height-age models |
|
WANG Ya-nan,WANG Jun-hui,QI Wan-yi,et al
|
| Abstract: |
| 【Objective】Parameters of height-age models for Pinus armandi Franch provenances were analyzed to improve the establishment of generalized height-age model for Pinus armandi Franch.【Method】Pinus armandi Franch in provenance test forest in Dalaoling Forest Farm,Yichang,Hubei were chosen as experimental materials.Regression accuracies of basic model,nonlinear mixed model and dummy variable model describing height growth of 11 Pinus armandi Franch provenances with stand ages of 4－10,24 and 32 years were compared.【Result】Accuracy of Richards equation was higher than those of Logistic equation,Schumacher equation and Korf equation to describe the basic height-age model for Pinus armandi Franch provenances.Both nonlinear mixed model and dummy variable model showed that there was significant difference in asymptote parameter but not in growth-rate parameter or shape parameter for 11 Pinus armandi Franch provenances.Averaged determination coefficient (R2),residual sum of squares (SSE),absolute residuals means (|E|) and minimum root mean square error (RMSE) of dummy variable model and nonlinear mixed model were 1.37% higher,18.27% lower,12.77% lower and 9.6% lower than those of the basic model.Dummy variable model was slightly better than nonlinear mixed model with increase of those four accuracies by 0.1%,1.8%,4%,and 0.9%,respectively.【Conclusion】Pinus armandi Franch provenances had significant effect on asymptote parameter but not in growth-rate parameter and shape parameter.The dummy variable model was suitable for the establishment of generalized height-age model for Pinus armandi Franch. |
| Key words: Pinus armandi Franch dummy variable model nonlinear mixed model provenances height-age models |
