浙江省毛竹竹秆生物量模型

Stem Biomass Models of Phyllostachys edulis in Zhejiang Province

【目的】通过样地实测获取毛竹竹秆生物量数据,研建基于不同变量的生物量模型并作比较分析,确定适宜的预测变量及模型,以精准估计毛竹竹秆生物量,为浙江省毛竹林立地质量评价和高效培育提供依据。【方法】从浙江省东、南、西、北、中不同区域选择10个县市采伐216株样竹,并进行样竹测量。引入胸径(D)、竹龄(A)和胸高竹节长(L)变量,利用全部样本信息,基于3个不同异速生长方程拟合竹秆生物量模型。采用似然估计法判定误差结构,确定模型拟合方法。通过3个模型的拟合优度及预估精度的比较分析,确定适用于浙江省的毛竹竹秆生物量模型。【结果】竹秆含水率逐年下降,Ⅴ度竹的平均含水率较Ⅰ度竹低24%;竹秆生物量占地上部分生物量比重逐年增加,且Ⅴ度竹占比超过80%;利用似然估计法分析确定生物量模型误差结构为乘积型,应采用对数转换的线性回归进行模型拟合;经检验,基于胸径的一元模型(M1)W=0.1046D^2.2578确定系数(Ra^2)仅为0.7742,而基于胸径-竹龄的二元模型(M2)W=0.0520D^2.2052A0.4457和胸径-竹龄-胸高竹节长的三元模型(M3)W=0.0265D^2.1439A^0.4495L^0.2629确定系数均达到0.89,且模型M3的估计值标准差(SEE)和平均系统误差(MSE)均为最小;3个对数回归模型在不同径阶范围的预估精度均较高,预估偏差接近于0,其中模型M3在不同径阶的预估效果均为最佳。【结论】由于模型校正后预估精度有所下降,故本研究在进行对数模型反对数转换时不作校正。二元和三元模型比一元模型具有更高的拟合优度和预估精度,确定基于胸径-竹龄-胸高竹节长的模型M3为最佳模型,即W=0.0265D^2.1439A^0.4495L^0.2629。

【Objective】The stem biomass of moso bamboo (Phyllostachys edulis) were accurately measured in sample plots. Proper prediction variables and models were determined on the basis of establishment and comparison among different biomass models with different variables. And the research is carried out to accurately estimate the stem biomass and provide a theoretical basis for the site quality assessment and efficient cultivation for bamboo forest in Zhejiang Province.【Method】 Firstly, mensuration of 216 sample bamboos harvested from 10 counties that distributed in eastern, southern, western, northern, and central part of Zhejiang Province was carried out. Secondly, the diameter at breast height (D), bamboo age(A) ,and internode length of bamboo at breast height (L) were introduced. Three different stem biomass models were fitted based on the allometric growth equations and all the sample information. Then, the model fitting method was selected by error structure that decided by the likelihood analysis. Finally, the most suitable stem biomass model was determined on the basis and analysis of the fitting goodness and prediction accuracy of the three different bamboo stem biomass models.【Result】The moisture content of bamboo stem decreased with years and the mean water content at the age of degree Ⅴ was 24% lower than that at degree Ⅰ. While bamboo stem biomass accounted for an increase in the proportion of above-ground biomass year by year and that at degree Ⅴ was more than 80%. The error structure of biomass models was determined to be multiplicative based on the likelihood analysis, thus the log-transformed model for fitting was required. 3) Upon accuracy inspection, the coefficient of determination (Ra^2) for model (M1) W=0.104 6D^2.257 8 was 0.774 2, lower than that of the binary model (M2)W=0.052 0D^2.205 2A^0.4457 based on D-A and the trigram model (M3) W=0.026 5D^2.143 9A^0.449 5L^0.262 9 based on D-A-L, whose value were up to 0.89. Meanwhile, the standard error of the estimate (SEE) and the mean absolute error (MAE) of model (M3)were the minimum. The three log-transformed models predicted well among different diameter classes as the prediction error were all close to 0. Over all the model M3 performed optimally among different classes. 【Conclusion】This study conducts the anti-log transformation of the log-transformed model without correction, for it can reduce the prediction accuracy. The binary and trigram models perform better than the unary model in the fitting goodness and prediction accuracy. Thus the optimum model is W=0.026 5D^2.143 9A^0.449 5L^0.262 9, based on the variable D-A-L.