Abstract:
Vegetation mapping is one of essential and fundamental topics of vegetation ecology study.Due to the complex and heterogeneity of mountain environments,vegetation mapping in mountainous region is a big challenge.This paper provides a brief introduction to the procedure of mountain vegetation mapping by using remote sensing.We focus on the current issues and practices of satellite image classification in mountain areas for improving the classification accuracy.This article suggests that ① in mountain areas,the classification accuracy can be significantly improved by training vegetation or land cover class with different spectral response patterns on different aspects or slopes;② many the topographic corrections models can significantly improve the accuracy of image classification;③ image classification for mountain areas,advanced non-parametric classifiers,such as neural networks,decision tree,or knowledge-based approach,should be the better choice.Fuzzy-set technique or spectral mixture analysis sub-pixel classification can be used to overcome the mixed pixel problem for the coarse and medium spatial resolution data.For high resolution data,object-based classification is the better choice;④ ancillary data,particularly DEM data can be combined with remote sensing data for improving image classification or vegetation mapping in mountainous regions.Advanced topographic normalization model and image classification algorithm should be developed in future.Integration of remote sensing and geographical information systems (GIS) appears as a new research topic.Currently,there exists a need to establish the link between theories and practices of ecological research strategies on one side and remote sensing image analysis on the other side.