Abstract: [Objective] Hickory(Carya cathayensis), one of the most important cash forests in Zhejiang province, plays an important role in improving economic conditions for local people and government. Currently, the hickory plantation area is mainly calculated from the estimation of hickory owners, but this area amount is often inaccurate and lack of spatial distribution information. Remote sensing with its unique characteristics in data collection and presentation has become the primary data source for mapping land cover distribution in a large area. However, mapping hickory plantation using remote sensing data remains a challenge because of the fact that hickory is a broadleaf tree and its plantation is often confused with other broadleaf forests in spectral signatures. Therefore, this research selected region of western Lin' an county, Zhejiang province, as a study area to explore the approach to map hickory distribution. Two Landsat 8 OLI images with leaf-on and leaf-off seasons in 2013 were used.[Method] Firstly, spectral mixture model (LSMM) was used to unmix Landsat multispectral imagery into three fraction images-green vegetation, shade and soil. Secondly, because hickory plantation has slightly different forest stand structure comparing with other broadleaf forest, their compositions of green vegetation, shade, and soil will be various. Based on this feature, three new indices, those are, vegetation-soil index, vegetation-shade index, and normalized multi-fraction index were proposed. Field survey data covering hickory plantations and other broadleaf forests were used to conduct a comparative analysis of these fraction images and newly proposed indices for the separation between hickory and other broadleaf forests. Thirdly, a decision-tree classifier was constructed by taking into account of Normalized Difference Vegetation Index (NDVI) and new index for mapping hickory distribution. Finally, the land-cover types of the research area were divided into two categories: hickory-others. The accuracy assessment of classification map was obtained by using field inventory data and high-resolution image of Google Earth.[Result] This study indicated that the normalized multi-fraction index could enlarge the difference of hickory from other broadleaf forests and could be successfully used to extract hickory plantation in this study area. The accuracy assessment result indicated that an overall accuracy of 88.67% with kappa coefficient of 0.76 was obtained in this study and implied that the LSMM based approach was promising in mapping hickory plantation.[Conclusion] Comparing with commonly used classification methods, the proposed normalized multi-fraction index has advantages in physical meaning, easy use and understanding, and the requirement in sample plots, thus, this new approach has the potential to provide a better classification accuracy than traditional classification algorithms. Furthermore, this approach may be used to map other plantations such as bamboo forest spatial distributions.

Key words: Landsat 8 OLI imagery, Garya cathayensis, linear spectral mixture model, normalized multi-fraction index