Built environment and taxi commuting carbon in Shanghai: volume, not intensity, and a definitional-coupling trap

Replication data and code for the study "Volume, not intensity: a geocomputational framework for attributing taxi commuting carbon to the built environment in Shanghai." This repository contains the grid-level analysis dataset (1,494 cells) and the complete code to reproduce the grid-level results: spatial econometric models (OLS, spatial lag/error, spatial Durbin, GWR), interpretable machine-learning attribution (XGBoost with SHAP and ALE), and counterfactual multi-objective optimisation (NSGA-II) with the decision-variable-independence check. Note: Raw taxi GPS trajectories and mobile-signalling job-housing data are access-controlled and cannot be redistributed. Trajectory-derived quantities (per-trip carbon accounting, commuting identification, deadheading, short-trip shares) therefore cannot be recomputed from this repository; the grid-level analysis on the shared aggregated data and code is fully reproducible. Manuscript submitted to Computers, Environment and Urban Systems.