Classification of dominant tree species in an urban forest park using the remote sensing image of WorldView-2

Abstract

There are many land use types and tree species in urban forest parks in which the human disturbance is frequent. Using remote sensing images to estimate the main tree species may provide a scientific basis for the making of sustainable management measures for scenic forest. In this article, Zijin Mountain National Forest Park in Nanjing, China, was selected as the case study area, and WorldView-2 data in December 2011 was chosen as the main information sources. Three kinds of band combinations were compared by using index of classification accuracy. Then the optimal combination was used to do supervised classification through three classification methods of decision tree classifier, neural networks, and support vector machine classification to distinguish the land use and the main species in the study area. The results showed that:1)The classification accuracy of 8-band combination of WorldView-2 is the highest and the overall accuracy and Kappa coefficients are 80.81% and 0.77, respectively, followed by the new 4-band combination and the standard 4-band combination. 2) Using the 8-band combination, the performance of decision tree classification is the best with overall classification accuracy of 87.10% and Kappa coefficient of 0.85, while the performance of neural networks classification is the worst with overall classification accuracy of 73.85% and Kappa coefficient of 0.70. 3) When comparing the accuracy of different tree species using decision tree classification, classification accuracy of the major local species is high, while the accuracy of foreign pine and cypress is relatively low.