哑变量在森林蓄积量模型估测中的应用

Application of dummy variables in estimation of forestry volume model

【目的】建立含哑变量的林分蓄积量估测模型,分析哑变量在香格里拉高山松林分蓄积量模型中的意义与作用。【方法】以香格里拉为研究区,基于2008—2009年3幅TM遥感影像与2008年抽样控制样地数据,对香格里拉高山松林分神经网络模型与考虑龄组构造的哑变量神经网络模型两种类型建立蓄积量遥感估测模型,并进行精度评价。对比模型的估测值与实测值,计算模型残差,检验各龄组残差均值与0之间的差异性;同时对模型的预测值结果进行组间均值的差异性检验,以此作为确定龄组分类形式构建哑变量的标准与依据。【结果】2个模型的独立样本检验结果表明,引入哑变量的神经网络估测模型比神经网络模型拟合效果要好,其决定系数要高于神经网络模型,决定系数从0.516提高到0.783。模型预估精度从神经网络模型的66.3%提高至哑变量模型的74.8%,估算误差优于神经网络模型。【结论】根据模型的残差差异性结果得出,哑变量模型可以在一定程度上解决在估测幼龄林、中龄林蓄积量低值高估的问题;可见引入哑变量估测森林蓄积量的方法是相对有效的。

【Objective】The stand volumeestimation model with dumb variables is established,which will analyze the significance and effect of dumb variables in stand volume model of Pinus densata forest in Shangri-la.【Method】Taking Shangri-la city of Yunnan as the research area,and using 3 TM remote sensing images(2008,2009)and sampling control sample data(2008),There are two types stand volume remote sensing estimation modelof neural network anddummy variable neural network considering the age groups of pinus densataforest in Shangri-la are established,and the accuracy of models is evaluated.Calculating the model residualsaccording to the estimated and measured values of the model,and testing the difference between the mean residuals of each age group and 0,Also,the difference between the means of group of the predicted values of the model was tested,which is used as the standard and basis for determining age group classification to construct dummy variable.【Result】The test results of independent sampleshow that the neural network model with dummy variable is better than the neural network model,and its decision coefficient is higher than the neural network model,the decision coefficient is increased from 0.516 to 0.783.The prediction accuracy of the model is improved from 66.3%of the neural network model to 74.8%of the dummy variable model,the estimation error of neural network model with dummy variableis better than the neural network model.【Conclusion】According to the result of the residual difference of the model,to a certain extent,the dumb variable model can be used to estimate the low valuation of young and middle aged forest stocks;It is obvious that the method of estimating the forest storage volume by the dumb variable is relatively effective.