Estimation of aboveground carbon stock of Moso bamboo (Phyllostachys heterocycla var. pubescens) forest with a Landsat Thematic Mapper image

Abstract

The extensive distribution of bamboo forests in South and Southeast Asia plays an important role in the global carbon budget. It is an urgent task to accurately and in good time estimate carbon stock within these areas. In this study, linear regression, partial least-squares (PLS) regression and backpropagation artificial neural network (BP-ANN) with a Gaussian error function as the activation function of the hidden layers (Erf-BP) were used to estimate aboveground carbon (AGC) stock of Moso bamboo in Anji, Zhejiang Province, China. Based on the combined use of Landsat Thematic Mapper (TM) and field measurements, the results indicate that the Erf-BP model provided the best estimation performance, and the linear regression model performed the poorest. This study indicates that remote sensing is an effective way of estimating AGC of Moso bamboo in a large area.

Acknowledgements

We acknowledge support from the National Natural Science Foundation (grant nos 30700638 and 30771715), the ‘948’ item of the National Forestry Bureau (grant no. 2008-4-49), the national ‘863’ program (grant no. 2006AA12Z104) and items of the Science and Technology Department of Zhejiang Province (grant nos 2007C13041 and 2008C12068) and thank the Anji Forestry Bureau for assistance during the field inventory. We also thank the reviewers for their constructive comments and assistance in improving this paper.