Monthly monitoring of urban development and renewal at the block-level in China using Sentinel-2 time series

Abstract

Urban renewal has been elevated to a national strategy in China, leading to rapid development and transformation of street blocks. However, monitoring construction events at high temporal resolution remains challenging due to the limitations of existing methods, which often struggle with noise interference and lack continuous monitoring capabilities. To address this, we propose Semantic Similarity Contrast-based Street Block Monitoring (SSC-SB), a method that leverages Sentinel-2 time series imagery for automated, high-frequency detection of street block development and renewal. By extracting deep semantic features with a pretrained encoder, SSC-SB analyzes similarity curves to identify development and demolition construction events. Applied to the Middle Yangtze River Basin (MYRB) urban agglomeration shows that SSC-SB achieves 90.4% spatial domain accuracy, with construction start and end date detection accuracies of 68.8% and 54.9%, respectively. Results indicate an increasing emphasis on urban renewal, as demolished street blocks outnumbered new developments for the first time in 2023, with Hunan Province leading in renewal efforts, where renewal blocks accounted for 41.5% of all changed street blocks, reflecting a balanced focus on expansion and infrastructure renewal. Transfer experiments in Xi’an further demonstrate that SSC-SB retains up to 80% of the performance of a locally trained model when applied across regions without fine-tuning, indicating a decent level of generalizability. By providing fine-grained, continuous monitoring, SSC-SB presents a scalable solution for tracking urban transformation.