云南松地上生物量模型研究

森林生物量作为森林生态系统的最基本数量特征,是研究许多林业问题和生态问题的基础,但由于地域的不同,地上生物量及各分项生物量存在差异。以西藏、云南2个省(自治区)的130株实测云南松Pinus yunnanensis生物量数据,分别用传统回归方法和利用引入地理区域为特征的哑变量方法建立了地上总生物量和地上各分项生物量的一元(胸径为自变量)、二元(胸径和树高为自变量)和三元(胸径、树高、冠幅为自变量)模型。结果表明:所建生物量模型中,地上总生物量模型精度最高,预估精度为0.9300~0.9600,其次是树干、树皮和干材生物量模型,预估精度为0.9000~0.9500,树叶生物量模型的预估精度相对较低,其值为0.8500~0.8900,而且所有的模型都满足二元模型的预估精度和确定系数比一元模型高,与三元模型相差不大。引入哑变量后的模型中,不管是一元模型、二元模型还是三元模型,模型的确定系数、预估精度都相应提高,确定系数为0.7300~0.9600,预估精度为0.8800~0.9600,而且估计值的标准误差和平均相对误差都减少了。因此,构建不同区域地上生物量和和各分项生物量模型时,建议引入哑变量,以提高模型精度和适用性,来解决不同地区模型不相容的问题。

An aboveground biomass model for Pinus yunnanensis

Forest biomass, the most basic quantitative characteristic of forest ecosystems, is the basis of many forestry and ecological problems. Because of regional differences, measurements for aboveground biomass and the biomass of six components, including stems, branches, crowns, foliage, boles, and bark, often differ. To determine if regional biomass characteristics introduced as dummy variables in one, two, and three variable biomass models effectively improved accuracy and R2 of the models, biomass data from 130 sampled trees of Pinus yunnanensis was determined for total aboveground biomass and biomass of components, including diameter alone; diameter and height; and diameter, height, and canopy width, as independent variables. Then traditional regression was used incorporating geographic areas as characteristics of dummy variables to develop one, two, and three variable biomass models in Xizang and Yunnan. Heteroscedasticity for each biomass model was eliminated with weighted regression. Results of total biomass models showed that the model for total aboveground biomass had the highest predicted precision (P) with 0.9300-0.9600 followed by models for stem, bark, and bole biomass with a precision of 0.9000-0.9500. Predicted precision for the foliage biomass models was relatively low, but it was still greater than 0.8500. The coefficient of determination (R2) and P for the two variable model compared to the one variable model was more greater, but was no differences in three variable biomass models. With the introduced dummy variable, R2 maximum reached 0.7300-0.9600 and P maximum reached 0.8800-0.9600, thereby reducing the standard error and the average prediction error of the estimated value. Therefore, when constructing aboveground biomass models for different regions, the dummy variable model should be used to improve accuracy and generality of the aboveground biomass model, thereby helping to settle incompatibility problems between models of different regions.