Dynamic monitoring and evolution of urban green space landscape sustainability based on spatiotemporal analysis algorithm

To address the issue of disconnect between the spatial-temporal characteristics of urban green space landscape dynamics and sustainability assessment, this study utilises the spatial-temporal DBSCAN (ST-DBSCAN) algorithm to identify hotspot areas of green space coverage changes and the spatial-temporal variations in green space fragmentation, and combines this with a geographically weighted regression (GTWR) model to quantitatively analyse the impact of driving factors such as population growth and land development intensity on the ecological service functions of green spaces.The experimental results show that the LSTM+ST-ConvNet model predicts the green space area for 2020-2024 with an error of only 50 square kilometres, outperforming other prediction models such as ARIMA, XGBoost, and Random Forest (the ARIMA model had an error of 70 square kilometres).This model can more accurately predict trends in green space area changes, providing theoretical support for urban green space.