Spatiotemporal Segmentation of Spaceborne Passive Microwave Data for Change Detection

Highly repetitive global-scale remote sensing systems, such as the Special Sensor Microwave/Imager (SSM/I), provide essential tools for monitoring changes on the Earth's surface. This letter presents a time-series segmentation and classification method to identify surface changes and to estimate the duration (days) for the changes using daily SSM/I observations. The method was developed based on a bottom-up segmentation algorithm for time-series data. The attributes of the linear segments provide the basis for understanding and classifying the surface changes. In the application examples, we calculated the number of surface snowmelt days at various locations on the Antarctic Ice Sheet by classifying the segmented time series of SSM/I brightness temperature observations. It is demonstrated that this novel method is robust to the data noise and efficient for processing large volume of spatially and temporally continuous remote sensing data for environmental monitoring.