From static supply to time-weighted path-based dynamic blue-green exposure: spatial mismatches between blue-green...

Assessments based only on blue-green resource supply within fixed spatial units may not capture environmental conditions encountered along movement routes or their spatial co-occurrence with platform-recorded use. Using Tianjin as the study area, we integrate publicly shared Keep walking trajectories with street-view imagery, remote-sensing imagery, environmental stress, and socioeconomic data. At the grid level, we measure static blue-green supply (SBG), time-weighted path-based dynamic blue-green exposure (DBG), environmental stress (ES), and Keep walking activity density (WAD). We then apply population-weighted Gini coefficients, quadrant-based mismatch analysis, a composite Priority Index (PI), and interpretable machine-learning methods with spatial cross-validation. Population-weighted Gini coefficients were 0.448 for SBG, 0.331 for DBG, and 0.374 for ES, indicating different degrees of distributional concentration relative to population. The low SBG-high WAD, low DBG-high WAD, and high ES-high WAD classes overlapped only partly, indicating distinct spatial patterns among resource supply, reconstructed-route exposure, environmental burden, and platform-recorded use. All four predictor domains provided predictive information, while out-of-fold SHAP patterns suggested broad nonlinear predictive associations with apparent turning points for several variables. Random Forest performance under spatial five-fold cross-validation was nevertheless limited ( $R^2$  = 0.051 ± 0.027), so these model-based patterns are treated as exploratory and complementary to the main spatial mismatch analysis. Static resource supply and reconstructed-route DBG therefore represent distinct environmental dimensions. Taken together, these measures provide a descriptive approach for distinguishing forms of environment-use mismatch and screening candidate areas for subsequent field verification.