西北林学院学报2024，Vol.39Issue(4)：30-38,9.DOI:10.3969/j.issn.1001-7461.2024.04.04
格氏栲天然林灌木生物量模型研究
Shrub Biomass Models of Castanopsis kawakamii Natural Forest
摘要
Abstract
In order to explore a suitable method to predict the biomass of shrubs of Castanopsis kawakamii natural forest,this study based on the measured data of four kinds of shrubs in Sanming C.kawakamii Nature Reserve,Fujian Province,and constructed a nonlinear independent model of the total biomass and each component biomass.The compatibility models were used to solve the biomass compatibility problem,and the weighted regression method was used to eliminate the heteroscedasticity of the models.Then the artificial neural networks were established,and the accuracy was compared with the compatibility models.The results showed that:1)the independent models of the total biomass and each component biomass of the four shrubs were generally more affected by the ground diameter factor.The binary models were better than the univariate models in accuracy,and R2 basically increased by more than 0.02.2)The compatibility models had little difference in accuracy from the independent models,and there was even a decrease in the accuracy of some component models.3)Compared with the compatibility models,the accuracy of shrub bi-omass prediction by artificial neural networks was significantly improved.The R2 of the total biomass and each component biomass model of artificial neural network mostly increased by more than 0.03 compared with the corresponding compatibility model,and the maximum improvement value was up to 0.21.Moreo-ver,it also had a good prediction effect on the models with low accuracy in the compatibility models.For the component whose R2 was lower than 0.6 in the compatibility model,the R2 of the artificial neural net-work model could be improved by more than 0.08.Generally,the compatibility models could solve the problem of biomass incompatibility;the prediction accuracy of artificial neural networks was higher than those of traditional models,and it was more worthy of choice when the performance of traditional models was mediocre for biomass prediction.The study provides theoretical reference for accurately predicting shrub biomass of Castanopsis kawakamii natural forest by conducting the comparison of two models.关键词
灌木生物量/非线性/相容性模型/人工神经网络/天然林Key words
shrub biomass/nonlinearity/compatibility model/artificial neural network/natural forest分类
农业科技引用本文复制引用
夏子濠,贾勃,王新杰,刘佳荣..格氏栲天然林灌木生物量模型研究[J].西北林学院学报,2024,39(4):30-38,9.基金项目
国家林业和草原局项目"森林质量精准提升监测理论与实证研究"(JYC2019-107). （JYC2019-107）
