基于生态因子与神经网络的杉木叶片碳氮磷含量预测

Prediction of Carbon,Nitrogen,and Phosphorus Contents of Chinese Fir Based on Ecological Factors and Artificial Neural Networks

[目的]利用神经网络所具有的输入层与输出层间存在的高度非线性映射关系,对杉木叶片C、N、P含量实现准确、经济、快捷的预测。[方法]以我国亚热带地区杉木人工林为研究对象,运用径向基函数(RBF)神经网络在杉木叶片C、N、P含量与地理、气候及土壤性质等生态因子间构建最优预测模型,并结合已发表文献数据进行叶片C、N、P含量预测。[结果]模拟预测叶片C、N和P含量分别为476.68、12.27和1.24 mg·g^(-1),其中N含量远低于我国陆地植物叶片平均含量;叶片C/N、C/P和N/P平均值分别为40.28、412.01和10.50。预测结果与实测值较为符合,表明RBF人工神经网络模型用于预测杉木叶片C、N、P含量与生态因子的关系是可行的。[结论]模型可以较为准确地估测杉木叶片C、N、P含量,平均误差分别为1.82%、9.88%和7.02%。较低的叶片N含量和N/P表明亚热带地区杉木生长主要受到N素限制。

[Objective]To achieve the accurate,economical and quick prediction of leaf carbon,nitrogen,and phosphorus contents of Chinese fir.[Method]Taking the Chinese fir(Cunninghamia lanceolata)plantations in subtropical China as objects,a RBF(radial basis function)neural network with highly nonlinear mapping relationships between input layer and output layer was used to build the optimal prediction models for the leaf C,N,and P contents of Chinese fir and ecological factors including geography,climate and soil properties.[Result]The simulation prediction of leaf average C,N,and P contents were 476.68,12.27,and 1.24 mg·g^(-1),respectively,the leaf N content of Chinese fir was far less than that of terrestrial plants in China;the leaf average C/N,C/P,and N/P were 40.28,412.01,and 10.50,respectively.The prediction results were well consistent with the measured values,indicating that it was feasible to use the RBF neural network model for predicting the relationships between leaf C,N,and P contents and ecological factors.[Conclusion]These models could accurately estimate the leaf C,N,and P contents of Chinese fir,the mean errors are 1.82%,9.88%,and 7.02%,respectively.Both the relatively low leaf N content and N/P indicate the growth of Chinese fir is limited by N element in subtropical China.