Mapping Forest Disturbance Types in China with Landsat Time Series

Forest is an important natural resource for the Earth. However, forest is frequently disturbance by different agents, which has different cascading impacts on surface energy balances, carbon dynamics et al. To better understand forest disturbance process in China, in this study, an object-based time series trajectory analysis technology was used to classify forest change data into different forest disturbance types, including forest logging and forest fire. The specific process includes: first, a hierarchical sample collection scheme was designed, and the visual interpretation of the samples is completed by different interpretation experts with the assistance of multi-source data; secondly, image segmentation of forest change data was performed, and the forest disturbance pixels that occur in the same year and are spatially adjacent are classified as the same object and used as the basic unit of subsequent processing; thirdly, the Landsat time series features of each image segmentation object were extracted, and the features representing their time series trajectory changes were further extracted; finally, the classifier was trained and optimized to complete the automatic classification of the national area. Experimental results show that the spatial and temporal distributions for different disturbance types vary greatly in China.