Seasonal Green view index mapping for 19 Chinese cities based on deep learning network

Multi-temporal mapping of the Green View Index (GVI) is essential for understanding how urban residents perceive seasonal variations in streetscape greenness. Compared to street view imagery (SVI), remote sensing data offers greater temporal resolution and broader spatial coverage, providing new opportunities for dynamic monitoring of urban greenery. However, existing GVI estimation methods rely heavily on SVI, making it difficult to achieve cross-city and multi-seasonal mapping. This limitation not only hinders accurate characterization of the spatiotemporal dynamics of GVI but also introduces potential errors and uncertainties in urban studies and related policymaking. To address these challenges, this study leverages multi-source remote sensing data and deep learning techniques to develop the Seasonal Green View Index-2023 (SGVI-2023) dataset, covering 19 major cities across China . During model training, we collected approximately 1.1 million paired samples of multi-source satellite data and SVI from 2019 to 2023 across the 19 cities and applied rigorous standards for data preprocessing and partitioning. Evaluation results demonstrate the high accuracy of SGVI-2023 in GVI estimation, achieving a Pearson correlation coefficient of 0.875 at the point scale and 0.93 at the street scale. To the best of our knowledge, SGVI-2023 is the first cross-city, seasonally resolved GVI dataset derived from remote sensing data. It offers a valuable data resource for dynamic monitoring of urban greenness from a human visual perspective and supports more refined, evidence-based urban planning and policy formulation.