广西杉木人工林多形地位指数模型构建

Polymorphous Site Index Models for Cunninghamia lanceolata Plantation in Guangxi Autonomous Region

以广西高峰林场杉木人工林为研究对象,使用森林资源二类调查数据中的树种因子与解析木数据,通过Richards方程与随机森林回归算法构建了两种多形地位指数模型。其中在利用随机森林回归算法构建模型时分别选取了地位指数和树高两种因子作为输出变量。结果表明:Richards方程模型与随机森林回归模型的决定系数(R^2)均在0.9以上,基准年龄下树高预测值与其对应的地位指数均方根误差小于0.3,说明两模型具有可靠性。经对比分析,两模型各具优势:Richards方程模型具有明确的表达式,相较于随机森林回归算法所构建的非参数模型,更适用于预测树木生长趋势,绘制树木生长曲线;随机森林回归模型比Richards方程模型误差更小,均方根误差(ERMS)降低了21%,平均绝对误差(EMA)降低了15%,且无需经过迭代运算,便可直接求解地位指数,克服了传统模型无法显式求解地位指数的缺点,求解效率更高。

Taking Cunninghamia lanceolata plantation in Gaofeng forest farm of Guangxi Autonomous Region as the research object,using the tree species factors and parse tree data in forest resource inventory and planning,the polymorphous site index model was constructed by Richards equation and random forest regression algorithm.The random forest regression algorithm was used to construct the model with site index and tree height as the output variables.The results show that R^2 of the two models are above 0.9,and the root-mean-square error between the tree height prediction value under the standard age and site index is less than 0.3,which indicates that the two models are reliable.Through comparative analysis,Richards equation model has a clear expression,which is more suitable for predicting tree growth trend and drawing tree growth curve than the nonparametric model constructed by random forest regression algorithm.Compared with Richards equation model,random forest regression model has lower error level.ERMS is reduced by 21%,E MA is reduced by 15%,and the site index can be solved directly without iterative operation.That overcomes disadvantages of traditional models that can not directly solve the site index and makes computing more efficient.