Evaluating the impact of environmental conditions on urban residents’ greenspace exposure during daily travel: An explainable machine learning approach

Urban greenspace (UG) is vital for urban climate regulation and public health, drawing increasing attention to greenspace exposure (GE) at different levels. However, limited understanding persists regarding human mobility-related GE, particularly the fine-grained dynamics of travel-related GE and the potential influence of environmental conditions such as weather and air pollution. This study examines how environmental conditions impact daily travel-related GE among urban residents, utilizing dockless bike-sharing data from Beijing, China. Firstly, spatiotemporal dynamics and inequalities in GE during travel were assessed using a population-weighted exposure model and the Gini index. Next, the effects of environmental conditions were evaluated through multiple models, including Ordinary Least Squares (OLS) regression and machine learning approaches: Random Forest (RF), Light Gradient Boosting Machine (LightGBM), Categorical Boosting (CatBoost), and Extreme Gradient Boosting (XGBoost). The deep learning network Long Short-Term Memory (LSTM) model was also included to account because of its effectiveness in processing time-series data. Model performance was evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), R-squared (R²), and cross-validation. Finally, SHapley Additive exPlanation (SHAP) and Partial Dependency Plots (PDP) were employed to analyze nonlinear effects and variable interactions. Results showed that XGBoost outperforms other models and is more applicable to small sample datasets than deep learning. Findings revealed that weather and air pollution significantly influenced GE during travel in addition to temporal factors (e.g., hour of the day, day of the week). Higher temperatures and lower humidity were associated with increased GE levels and reduced inequality. Severe ozone pollution events reduced GE levels but also lowered inequality. No significant impact of particulate matter (PM) on GE was observed due to the absence of severe haze events during the study period. These findings provide valuable insights for urban greenspace planning and strategies to promote healthy travel behaviors.