Influence of different topographic correction methods on the remote sensing extraction of Robinia pseudoacacia distribution

Robinia pseudoacacia is one of main tree species of environmental construction in loess hilly gully region, and topographic shadow is a main factor affecting the accuracy of its information extraction by remote sensing. A case study was carried out in Ansai County, Shaanxi Province of northwestern China, which belongs to the loess hilly gully region to compare different topographic correction methods in extracting information on the distribution of Robinia pseudoacacia. The topographic effects of the Landsat8 OLI(July, 2015) were removed by 5 common topographic models (Cosine, SCS, Minnaert, C, SCS+C), and the example-based feature extraction was used to extract the distribution of R. pseudoacacia. The results were evaluated by visual comparison, regression analysis and ground-based validation. Results show that: 1)all topographic correction methods could reduce the topographic shadow on the remote sensing image and both the Cosine model and SCS model over-corrected the shade area's image. 2) All topographic correction methods changed the mean and variance of each bands' radiance, and SCS+C model reached the expected goal. 3) Minnaert model, SCS+C model and C model decreased the absolute value of two regression model parameters(slope m and coefficient r) between all bands' radiance and cosi, however Cosine model and SCS model increased the absolute value of two regression model parameters in some bands. 4)All topographic correction methods could correct the image and decrease the omission of the extraction of R. pseudoacacia's distribution to a certain degree, but Cosine model even decreased user's accuracy by 2.47%. Minnaert model, SCS+C model and C model could improve both the producer's accuracy and user's accuracy, and the C model, considered as the best method in the research, increased the producer's accuracy and the user's accuracy by 16.66% and 2.82%, respectively. 5) All topographic correction methods increased the Kappa coefficient, and the Kappa coefficient was highest (0.76) after C model's correction. The result suggested that topographic correction was necessary to the extraction and the model chosen should take the condition of the research area into account, which provided an important basis for choosing a proper topographic method in the process of extracting R. pseudoacacia's distribution in the loess hilly gully region.