Global Urban Vegetative Cooling: City Boundaries and Analysis Tables, 1990–2024

This dataset accompanies the manuscript Global Weakening of Urban Vegetative Cooling and provides everything needed to reproduce the core analyses. It includes (1) a GeoPackage of 311 global city boundaries and (2) analysis-ready tables (CSV) with city-level means/medians, temporal trends, and standardized design matrices used for LASSO and partial-dependence diagnostics. Contents City boundaries: city_boundaries.gpkg (one polygon layer per city; join key = City). Tables (CSV): means.csv – long-term means (1990–2024) of responses & predictors. means_standardized.csv – same variables, z-scored; missing values imputed by column mean (for LASSO/PDP workflows). medians.csv – long-term medians (1990–2024). medians_standardized.csv – z-scored + imputed version of medians. trends.csv – per-city OLS slopes (1990–2024) for time-varying variables. trends_standardized.csv – modeling table with TREND_* (1990–2024 slope) and MEAN_* (1990–2024 mean) columns; z-scored; missing values imputed. trends_pvals.csv – p-values from per-city temporal regressions (1990–2024).See included README for details. Key variables Responses: mean_NDVI_raw (unitless), mean_LST_raw (°C), VegetativeCooling (slope of LST~NDVI; °C per NDVI; more negative = stronger cooling). Cumulative precipitation: median_precip_cumNN = median cumulative precipitation for warm-season months using the current plus N previous months (per city, 1990–2024). Aridity index (AI): AI_mean (mean pr/PET) and AI_trend (trend of pr/PET). Latitude_ABS: absolute latitude (|Latitude|). Static vs dynamic: Static predictors (e.g., elevation) appear only as MEAN_* (no trend). Notes on standardized tablesAll non-categorical columns are z-scored. Missing values were imputed by the column mean only to support complete-case modeling (LASSO and PDP). If you need coefficients/intercepts in original units, use the unstandardized means.csv/medians.csv. Elevation caveatElevation primarily derives from SRTM DEM, which covers ~60°N to 56°S. For far-north cities outside SRTM, values were filled from external sources; see README. Provenance & sourcesPredictors were derived from public, trusted repositories via Google Earth Engine (Landsat C2 L2, TerraClimate, ERA5-Land, GHSL, GPW, Yale YCEO SUHI, SRTM). Warm seasons use May–Sep (NH) and Nov–Mar (SH), 1990–2024. See README for dataset IDs and processing notes. UsageJoin tables to polygons on City. Use standardized tables for feature selection and PDPs; use unstandardized means/medians if you want effect sizes in physical units of the response. Related materials Code repository (GEE + MATLAB) and full workflow are documented in the project README. When you archive the code on Zenodo, link that record here under “Related/Alternate identifiers.” Suggested license Code: MIT (in code repo). Data derivatives: CC BY 4.0.Respect original data-provider terms (USGS, JRC, ECMWF, CIESIN, etc.). Keywords:urban climate; NDVI; land surface temperature; vegetative cooling; urban heat island; Landsat; TerraClimate; ERA5-Land; GHSL; SUHI; cumulative precipitation; aridity index; Google Earth Engine; MATLAB; global cities; geospatial