Spatial-temporal patterns of urban land use efficiency in china's national special economic parks

The effective utilization of urban land plays a significant role in achieving the 11th goal focuses on Sustainable cities and communities. China's national special economic parks (NSEP) have undergone a remarkable process of urbanization over the past 40 years, and the land for future development is becoming increasingly scarce. In this study, a convolutional neural network with attention mechanisms was proposed for extracting built-up area boundaries of NSEP; Additionally, the urban land use efficiency (ULUE) was estimated by combining multi-source nighttime light image and statistical data from yearbooks. Furthermore, we integrate a data-driven model for predicting ULUE. The empirical research involves a comprehensive analysis of the spatial characteristics of urban land expansion by using the Taylor index and spatial analysis. This study reveals the following findings: (1) Convolutional Neural Networks based on attention mechanisms can achieve high-precision extraction of urban boundaries with a Kappa value exceeding 0.92; (2)There is a significant positive correlation between the brightness of nighttime light imagery and the GDP of the secondary and tertiary industries; (3) ULUE exhibits evident dependence on urban scale; (4)The density-nuclear density curve features of ULUE with the peak gradually decreasing, which indicates a continuous expansion of the distribution range of land use efficiency. This research framework aims to provide the support for the scientific management and sustainable utilization of urban land while serving as empirical reference for future policy formulation and planning.