Search

Taking the moso bamboo production areas Lin'an, Anji, and Longquan in Zhejiang Province of East China as study areas, and based on the integration of field survey data and Landsat 5 Thematic Mappr images, five models for estimating the moso bamboo (Phyllostachys heterocycla var. pubescens) forest biomass were constructed by using linear, nonlinear, stepwise regression, multiple regression, and Erf-BP neural network, and the models were evaluated. The models with higher precision were then transferred to the study areas for examining the model's transferability. The results indicated that for the three moso bamboo production areas, Erf-BP neural network model presented the highest precision, followed by stepwise regression and nonlinear models. The Erf-BP neural network model had the best transferability. Model type and independent variables had relatively high effects on the transferability of statistical-based models. [ABSTRACT FROM AUTHOR]

Taking the moso bamboo production areas Lin'an, Anji, and Longquan in Zhejiang Province of East China as study areas, and based on the integration of field survey data and Landsat 5 Thematic Mappr images, five models for estimating the moso bamboo (Phyllostachys heterocycla var. pubescens) forest biomass were constructed by using linear, nonlinear, stepwise regression, multiple regression, and Erf-BP neural network, and the models were evaluated. The models with higher precision were then transferred to the study areas for examining the model's transferability. The results indicated that for the three moso bamboo production areas, Erf-BP neural network model presented the highest precision, followed by stepwise regression and nonlinear models. The Erf-BP neural network model had the best transferability. Model type and independent variables had relatively high effects on the transferability of statistical-based models.