• 研究报告 •

### 山西阳泉地区乔木林地上碳密度遥感估测

1. (1山西大学环境与资源学院， 太原030006； 2山西省林业调查规划院， 太原 030012)
• 出版日期:2014-09-10 发布日期:2014-09-10

### Estimation of aboveground carbon density for tree forests based on remote sensing data in Yangquan of Shanxi Province, China.

LI Jiao1, ZHANG Hong1**, ZHANG Li-qiu1, HAN Jian-ping2

1. (1College of Environment and Resource Sciences, Shanxi University, Taiyuan 030006, China; 2 Shanxi Institute of Forestry Inventory and Planning, Taiyuan 030012, China)
• Online:2014-09-10 Published:2014-09-10

Abstract:

In order to investigate the feasibility of using remote sensing data to determinate the aboveground carbon density for tree forests, we estimated the biomass and carbon density of forests in Yangquan region of Shanxi Province by using the variable BEF (biomass expansion factor) method based on the forest inventory data. We then selected the NDVI, RVI, bands of SPOT images and environmental factors (elevation, slope, aspect etc.) as independent variables, and established an estimation model by using the enhanced BP neural network method to derive a distribution map of the carbon density. The biomass of tree forests in Yangquan was 552774 t, and the carbon density was 11.38 t·hm-2. Needleleaved forest, young forest and artificial forest had the largest biomass, while broadleaved forest, mature forest and natural forest had the largest carbon density. Model predictions of carbon density were successful, with the average relative errors and the average absolute values of relative error of simulation results for needleleaved forest, broadleaved forest, and mixed forest being 2.40%, 6.87%, -4.09%, and 6.83%, 2.77%, 3.99%, respectively. The simulation accuracy of tree forest distribution map derived by the enhanced BP neural network model was 85.05%, implying that artificial neural networks developed a new idea for fast and accurate estimation of forest carbon density, providing a scientific basis for future survey and management of forest resources.