An ensemble method for monitoring land cover changes in urban areas using dense Landsat time series data

Abstract

This study proposes a new method for monitoring land cover change in urban areas using all available Landsat time series data, named the Ensemble of Bidirectional Time Series Analysis (EBTSA). In this method, the bidirectional Continuous Change Detection and Classification (CCDC) and the Chow Test are combined to improve the robustness against data scarcity in earlier times and reduce break detection errors and refine classification results. There are three key stages in this method: break detection using bidirectional CCDs, break refinement using the Chow Test, and bidirectional model integration and classification. The EBTSA method was evaluated over the Tianjin metropolitan area in China using Landsat data from 1989 to 2018. Results show that the proposed method improved spatial and temporal accuracies of both land cover classification and change detection, by reducing the influence of sparser Landsat data in the earlier years and the break detection errors. Using the land cover change results in the Tianjin area obtained using the EBTSA method, we analyzed the spatio-temporal distribution of land cover classification and change detection. It is found from the results that the Tianjin area experienced dramatic urban land changes, characterized by rapid urban expansion and noticeable transition from vegetation to urban land, with diverse changes from urban land to various nonurban land cover types. Results of this study showcase the effectiveness of the proposed method in land cover change monitoring in urban areas, which facilitates a more comprehensive understanding of urban land change dynamics and diversity.