A Global, Standardized City Segment Morphological Deprivation (CSMD) Model: Preprocessing, Training, Predictions, and Cross-Dataset Comparisons
Domain:
geospatialsocioeconomicRecord type:
This Zenodo record archives the reproducibility data package for the City Segment Morphological Deprivation (CSMD) model, a globally standardized framework for mapping morphologically deprived urban areas at neighborhood scale across 5,132 cities in 103 countries across Africa, Asia, and Latin America and the Caribbean.
The CSMD workflow integrates harmonized built-environment indicators from City Segments v1 with IDEABench reference labels to train a supervised Random Forest classifier. Internal validation is performed using a Leave-One-City-Out (LOCO) cross-validation framework, and the final production model is trained on 100% of the labelled data prior to global deployment. The trained model is subsequently applied to City Segments v1 to generate global segment-level deprivation predictions. The repository also includes comparative analyses with three external spatial datasets: the Slum Severity Index (SSI), Million Neighborhoods (MN), and WRI Intra-Urban Land Use.
This version updates the reproducibility package following final repository validation. It includes corrected WRI comparison summary outputs generated from preprocessed per-block WRI files, regenerated combined comparison figure outputs, revised coverage/omission outputs, and a SHA256 file manifest for the uncompressed data package.
Contents of this Zenodo package
The package includes the large data products and model artifacts that are intentionally excluded from GitHub:
-
models.zip— serialized final Random Forest model (rf_final_model_full.joblib). -
predictions.zip— per-country CSMD prediction GeoPackages (*_rf_preds.gpkg). -
ssi_clipped.zip— clipped SSI rasters aligned to CSMD city-segment coverage. -
mn_blocks_by_country.zip— per-country Million Neighborhoods blocks used in the comparative analysis. -
mn_comparison_files.zip— per-country MN-to-CSMD segment-level comparison files. -
wri_per_country_outputs.zip— preprocessed WRI per-country outputs used for WRI comparison metrics and final summary tables. -
GHS_STAT_UCDB2015MT_GLOBE_R2019A_V1_2_with_GHSPOP2023.gpkg— derived GHS-UCDB 2019 polygon file with GHS-POP R2023A 2025 city population estimates. -
zenodo_file_manifest.csv— SHA256 checksum manifest for all uncompressed files in the localdata_external/zenodo/package.
Recommended use with the GitHub repository
The corresponding code, documentation, small tables, selected figures, environment files, and workflow notebooks are available at:
To reproduce the repository workflows, download the ZIP files from this Zenodo record and extract them into the GitHub repository under:
data_external/zenodo/
After extraction, the folder should have the following structure:
data_external/zenodo/
├── models/
├── predictions/
├── ssi_clipped/
├── mn_blocks_by_country/
├── mn_comparison_files/
├── wri_per_country_outputs/
├── GHS_STAT_UCDB2015MT_GLOBE_R2019A_V1_2_with_GHSPOP2023.gpkg
└── zenodo_file_manifest.csv
The repository documentation, especially README.md, docs/data_availability.md, and zenodo/ZENODO_CONTENTS.md, describes how these files are used by the preprocessing, modelling, prediction, figure-generation, revision2 coverage/omission, and comparative-analysis workflows.
Data access and third-party datasets
This package contains derived and processed data products needed for reproducibility. Raw third-party datasets remain with their original providers and should be obtained from their official sources where required. These include City Segments v1, IDEABench v2, GHSL UCDB/GHS-POP, SSI source data, Million Neighborhoods, and the WRI Urban Land Use dataset.
IDEABench is subject to its own access conditions. Requests regarding IDEABench should be directed to the IDEABench data providers. Requests regarding the CSMD model, derived outputs, and repository workflows can be directed to Sai Ganesh Veeravalli.
Citation
Veeravalli, S. G., Blei, A. M., Friesen, J., Tareke, B., Kuffer, M., Persello, C., Maretto, R. V., Abascal, A., Georganos, S., & Thomson, D. R. The Hidden Burden of Morphological Deprivation in Small and Medium Cities. Accepted in principle, Nature Cities.