Large-scale forest resource dynamic monitoring using worldview-2 data - A case study in Pingtan Island, Fujian Province, China

Abstract

Due to its capacity to describe detailed information., high spatial resolution remote sensing technology has been used increasely in forest resource dynamic monitoring, especially at large scale. Based on 0.5m resolution Worldview-2 data and 1:10000 forest stand-wise vector data, monitoring on forest resource dynamic during 2010-2011 in Pingtan Island is carried out. Object-oriented classification method was adopted. Firstly, multi-resolution segmentation was done on Worldview-2 image, which produced two levels of objects. The boundary of objects at the first level corresponded to boundary of individual polygon in the forest stand-wise data coverage. The more homogeneous objects at the second level were smaller and within the objects at the first level. Then, hierarchical supervised classification was used to classify the second-level objects. The objects were assigned to seven different thematic layers according to the attribute of land use types of forest stand-wise data, then object samples were selected for supervised classification to identify the change of objects in each layer. Changed objects larger than 0.0667hm 2 were preserved and used for dynamics analysis. The classification accuracy was assessed by field survey and one-by-one visual interpretation. The result indicated that the overall accuracy was 90.85%. Forest in Pingtan Island had changed greatly during 2010-2011, area of forest increase and decrease is 2314.7817 hm2 and 1089.6508 hm2 respectively.