Data supporting the publication: CA-Based Simulation for Multi-Class Land Use Patch Succession in Urban Historic Areas: Mechanisms and Spatial Prediction

This dataset of historical land use change and driving factors in the Xinqiao Historic Area is constructed to serve the empirical research on micro-scale (10m×10m) multi-class land use patch succession in urban historic areas under rapid urbanization, with the core research objectives of exploring the spatial evolution mechanisms of historic area land use, conducting accurate spatial prediction and development vulnerability assessment, and addressing the bottlenecks of traditional Cellular Automata (CA) models such as spatial scale mismatch, rigid time steps and oversimplified neighborhood effects in historic area research, further providing quantitative data support for historic area conservation and short-, medium- and long-term urban spatial planning. Belonging to the empirical research field of urban geography and urban-rural planning, the research takes the 10 km² Xinqiao Historic Area in Bao’an District, Shenzhen as the case study area, and adopts a multi-source data collection method: integrating 1969–2023 multi-temporal satellite imagery (1–5 m resolution) from EarthExplorer, Google Earth Pro and Tianditu, which is processed through geometric correction, radiometric correction, cloud removal and bilinear interpolation resampling to 1 m resolution with a land use recognition accuracy of ≥95%; combining field surveys, the Baonan County Chronicle, genealogies and oral history accounts to reconstruct and vectorize the land use patterns and key facility elements of each period; extracting topographic features (30 m DEM resampled to 10 m) and Euclidean distances to natural, transportation and cultural facility elements, defining the neighborhood with a 3×3 grid and quantifying 10 types of land use features via one-hot encoding, with the collected data split into 80% training and 20% test sets for model validation. The dataset mainly consists of multiple Excel format files including multi-period land use conversion matrix data, spatial attribute data of 10 land use types (historic land, river, new residential land, etc.), quantitative data of driving factor feature variables (topography, spatial distance, neighborhood agglomeration, etc.) and land use transition probability simulation results, matched with a JSON format script for data reading and processing, all based on the 10m×10m grid units of the Xinqiao Historic Area, and the data can directly support the construction, training and validation of the Multi-class Land Use Patch Succession Cellular Automaton (MC-LUPS-CA) model and related land use simulation research.