Code, trained models, derived data, and simulation outputs for "An integrated simulation of urban expansion and regeneration based on CNN-LSTM and cellular automata"

This repository contains the custom code, trained models, derived data, and simulation outputs used in the study “An integrated simulation of urban expansion and regeneration based on CNN-LSTM and cellular automata”. The repository supports a spatiotemporal coupled cellular automata framework integrating CNN-LSTM-based suitability estimation with patch-based allocation for simulating urban expansion and urban regeneration in Wuhan, China. Files included in this repository are as follows: - code.7z: original source code of the modelling framework, including three CA models (a coupled CA model, a decoupled CA model, and a coupled CA model for projection to 2035). The folder “renewal_generate” is used to generate annual land-use maps containing urban regeneration and expansion, as well as the dynamic driving factors required by the models. The folder “LSTM_Trainer” contains the sampling and training code for the CNN-LSTM module.- cnn_lstm_model.zip: three trained CNN-LSTM models used in this study.- Performance_Evaluation.7z: code for model performance evaluation, including Figure of Merit (FoM) and Landscape Error (LE).- result.7z: simulation results under different iteration settings and scenario simulation results.- road_networks.7z: road network data derived from OpenStreetMap and processed through topology checking.- poi.zip: point-of-interest (POI) data obtained from the Amap platform; some layers were temporally matched where relevant time information was available.- static.7z: static factors and auxiliary files, including terrain, water bodies, restricted-area maps, parcel centroid records, and parcel neighbourhood records used to support and accelerate model computation. This repository does not include original source data that are subject to third-party licensing or terms-of-use restrictions. In particular, original land transaction records obtained from the China Land Market Network and original raw outputs obtained from the Amap platform are not redistributed. Users should obtain these data independently from the original providers and process them following the procedures described in the associated article. To fully reproduce the workflow, users also need to independently obtain additional external datasets, including GAIA annual impervious surface data, Wuhan parcel footprints (Gong, Chen, et al., 2020), and LandScan population data, and preprocess them to the same spatial extent, resolution, and raster dimensions as the modelling inputs used in this study. Some file paths specified in the code are environment-dependent and should be adjusted by users according to their own local directory structure and computing environment before execution. The repository is currently under continued curation, and its organisation and documentation may be further improved following publication of the article. The deposited materials are provided for research and verification purposes. Please cite the associated article when using the code, derived data, or simulation outputs contained in this repository.