To effectively and objectively evaluate the spatial efficiency of rail transit station areas, seventeen typical rail station areas in Xi’an were selected as the research object. An evaluation system for spatial efficiency was constructed based on data from field research, satellite images, Baidu heat maps, and station passenger flow statistics. Key factors such as land use, transportation systems, social aspects, and spatial efficiency are considered in the framework. A data envelopment analysis (DEA) method was used to evaluate the spatial efficiency of these sample station areas. The results are as follows. ① An incomplete symmetric relationship exists between the Constant Returns to Scale Technical Efficiency (Crste) and the Variable Returns to Scale Technical Efficiency (Vrste) of station area spatial efficiency. The keys to improving station area spatial efficiency include reducing redundant resource investments and establishing a rational resource allocation structure. ② For high-efficiency station areas, the Crste and Vrste are relatively high, with an overall increasing return to scale efficiency (Scale). In medium-efficiency station areas, the Crste is relatively high, but either Vrste or Scale is low. In low-efficiency station areas, the Crste is moderate, and both Vrste and Scale are low. The findings provide a reference for the intensive use of land around Xi’an rail stations, as well as support for the sustainable operation of rail transit.

For the intensive, efficient, and sustainable utilization of land resources, it is of great significance to optimize the spatial allocation of different types of land use intensity in rail transit station areas. The current land use optimization model has some shortcomings in objective function, constraint conditions, and the solution process. In response to this, a new multi-objective optimization model for rail transit station land use was built. With station space efficiency as the starting point, the three objectives of the model optimization were the traffic volume, environment quality, and land balance of the rail transit station, and the constraint conditions were the plot ratio, environment quality, and efficiency level. Lingo was used to solve the optimal plot ratio of different types of land use intensity. Compared with the non-inferior solution of the rail transit station area multi-objective original model, the ideal plot ratio of various land uses obtained by the optimized new model was more reasonable. There was a relatively large gap between the non-inferior solutions of some original models and the actual conditions. In contrast, the optimized new model had stronger maneuverability. The deviation ranges of the two models were −0.4% to 0.9% on the residential land plot ratio adjustment index, −3.2% to 4.8% on the public land plot ratio adjustment index, and 1.1% to 1.9% on the commercial land plot ratio adjustment index. This research aimed to provide a basis and reference for the land use and planning of Xi’an rail transit station.

Existing studies often examine associations between the built environment (BE) and low-carbon mode choice using aggregated travel outcomes, with limited attention to how these relationships vary across trip purposes. Using a final analytical sample of 855 respondents from Xi’an, China, this study integrates random forest (RF), SHapley Additive exPlanations (SHAP), partial dependence plots (PDPs), principal component analysis (PCA), and structural equation modeling (SEM) to examine purpose-specific BE importance, nonlinear response patterns, and perception-mediated pathways across six trip purposes. These include commuting, education, maintenance activities involving shopping and routine daily services, healthcare, recreation involving cultural, fitness, and entertainment activities, and outdoor leisure involving parks, squares, and scenic areas. The results show substantial heterogeneity across trip purposes. RF models achieved area under the receiver operating characteristic curve (ROC-AUC) values ranging from 0.742 to 0.861, while the most important BE dimensions differed across purposes. Nonlinear associations also varied in form and range across trip purposes, including an apparent turning point at approximately 200 m for the distance to the nearest bus stop in commuting trips, beyond which the predicted probability of low-carbon travel declined. SEM results further showed that perceived BE mediated only specific BE–travel pathways, with partial mediation identified for PCA3 in maintenance trips and PCA2 in recreation trips after accounting for community-level clustering. No full mediation was identified. These findings show that BE–travel relationships vary in environmental importance, nonlinear response, and perception-mediated pathways across daily activities. The results can help planners identify purpose-specific environmental priorities and develop more targeted community-scale interventions for sustainable transportation.

Cycling accessibility is a key indicator of urban resource equity and built environment performance. However, its relationship with equity, the importance of built environment factors, and nonlinear effects across facility types remain underexplored. This study combines the Gini coefficient with a generalized additive model (GAM) to examine the nonlinear accessibility–equity relationship and uses Light Gradient Boosting Machine (LightGBM) with SHapley Additive exPlanations (SHAP) to assess the importance and nonlinear effects of built environment variables. Potential effects of vegetation (Normalized Difference Vegetation Index, NDVI) and precipitation are also explored. Results show that both accessibility and equity peak in urban cores and decline toward peripheral areas. Facility types exhibit distinct patterns: accommodation services have high accessibility but low equity, life services require higher accessibility to achieve equity, and public services show an N-shaped relationship. SHAP analyses indicate that traffic signals, subway station accessibility, and street enclosure ratio are associated with higher cycling accessibility, whereas a higher sky view index is associated with lower accessibility. Observed moderating effects of vegetation and precipitation provide further insights into potential interactions. These findings offer guidance for improving cycling infrastructure and promoting more equitable and sustainable urban accessibility.