ABSTRACT
Mapping dynamics of impervious surfaces is important for analysis of urban environments, including biogeochemical cycles, climate change, and biodiversity. This study developed a new method for differentiating impervious surfaces from pervious surfaces based on temporal features. In this study, dynamic time warping (DTW) distance was introduced to measure similarity of temporal features from reconstructed time series Biophysical Composition Index (BCI). Then, semi-supervised Support Vector Machine (SVM) was employed to map dynamics of impervious surfaces by classifying temporal features at an annual frequency. Wuhan city was used as a case study to validate the performance of our method for the period from 2000 to 2015. In the 16 years period, impervious surfaces of Wuhan city increased from 515.27 km2 in 2000 to 2308.91 km2 in 2015, with an annual growth rate of 21.76%. Our findings showed that temporal domain had the potential to characterize features of different land cover, and our method also showed promise in differentiating impervious surfaces from pervious surfaces using temporal features instead of only spectral features.
