How coarse analysis can obscure key trends - a multi-scale study using very high-resolution intra-urban land data

Land use change (LUC) mechanisms are vital for understanding change processes and policy formulation. However, most previous studies neglected intra-urban land use evolution and the influence of spatial resolution. This study designed a multi-scale LUC driver analysis framework to analyze the drivers of intra-urban land use expansion in selected cities in Hunan Province, China, and explored the impact of spatial resolution on driver analysis. First, very high-resolution intra-urban land use survey data were used to explore the driving mechanisms based on Random Forest and Spearman’s correlation analysis at 2 m resolution. Then, results from other resolutions (5 m, 10 m, 30 m, 60 m, 90 m, and 120 m) were compared to the 2 m resolution. A scale effect assessment strategy based on Spearman’s correlation analysis, Sen’s slope, and Mann-Kendall (MK) trend test was devised to analyze the patterns of change in the driver mining results with spatial resolution. Findings reveal that night light brightness, elevation, residential density, population density, distance to transportation facilities, and distance to industrial parks predominantly drive intra-urban land use expansion, with different categories emphasizing specific factors. Analyzing intra-urban land use expansion mechanisms at low spatial resolutions (below 30 m) can significantly distort results and render the ranking of driving elements unstable. The loss of patch information due to reduced spatial resolution is the primary cause of distorted driving factor analysis results. It is recommended that related studies use data with a resolution of 30 m or higher. These findings offer valuable insights for optimizing land use policies and serve as a reference for selecting appropriate data resolutions in future studies.

土地利用变化（Land use change, LUC）机制是理解变化过程与制定政策的重要基础。然而，以往多数研究忽视了城市内部土地利用演化过程，以及空间分辨率的影响作用。本研究构建了多尺度LUC驱动因子分析框架，以中国湖南省选定的部分城市为研究对象，剖析其城市内部土地利用扩张的驱动机制，并探究空间分辨率对驱动因子分析的影响。首先，本研究以2米分辨率的超高分辨率城市内部土地利用调查数据为基础，借助随机森林（Random Forest）与斯皮尔曼相关分析（Spearman’s correlation analysis）探究其驱动机制。随后，将5米、10米、30米、60米、90米及120米等其他分辨率的分析结果与2米分辨率的结果进行对比。本研究提出了一种基于斯皮尔曼相关分析、森斜率（Sen’s slope）与曼-肯德尔（Mann-Kendall, MK）趋势检验的尺度效应评估策略，用以分析驱动因子挖掘结果随空间分辨率变化的规律。研究结果表明，夜间灯光亮度、海拔高度、住宅密度、人口密度、距交通设施的距离以及距工业园区的距离是驱动城市内部土地利用扩张的主要因子，且不同土地利用类别对应着各自的核心驱动因素。采用低于30米的低空间分辨率数据开展城市内部土地利用扩张机制分析，会显著扭曲分析结果，导致驱动因子的排序不稳定。空间分辨率降低导致的斑块信息丢失，是驱动因子分析结果出现偏差的主要原因。本研究建议相关研究采用分辨率不低于30米的数据开展相关研究。本研究结果可为土地利用政策的优化提供有益参考，同时也可为未来研究中合理选取数据分辨率提供借鉴依据。