Retrieval of forest growing stock volume by two different methods using Landsat TM images

Abstract

Forest growing stock volume (GSV) is one of the most important indicators in the field of forest resources investigation and monitoring. This article describes the application of two different methods, the multiple stepwise regression (MSR) model and the back-propagation neural network (BPNN), to retrieve forest GSV using Landsat Thematic Mapper (TM) images and field data. The article describes the data used, the retrieval methods adopted, and the results achieved. The results show that the surface reflectance of six bands significantly correlated with forest GSV, as did six vegetation indices, factors from principal component analysis and tasselled cap transformation, and three terrain factors. Moreover, texture features including Band 1_mean, Band 2_mean, and Band 3_mean were highly correlated with forest GSV. An optimal MSR model that included three factors was established for retrieving forest GSV using 53 remote-sensing factors. Three factors were included in the model. Leave-one-out cross-validation demonstrated that the model worked well. Finally, BPNN was constructed and the predicted result was highly consistent with measured forest GSV. In a comparison of the retrieved results with the MSR model and BPNN, the MSR model was better at quantitatively finding the correlation between each remote-sensing factor and forest GSV, and a linear equation could be acquired. However, BPNN was better at predicting forest GSV based on the field data. Additionally, the retrieved map of forest GSV for the whole study area by BPNN was much more consistent with the Landsat TM false-colour composite than that retrieved by the MSR model.

Acknowledgements

The authors thank the Academy of Forestry Inventory and Planning, State Forestry Administration, PR China, for providing the forest inventory data. We also wish to thank all those who have helped with the article.

Funding

This article was supported by the National Natural Science Foundation of China [grant number 41171330]; the National High Technology Research and Development Programme of China (863 Program) [grant number 2013AA12A302]; the Evaluation of Wetland Ecosystem.