Revealing the diurnal spatiotemporal dynamics and drivers of urban traffic congestion resilience

Urban traffic congestion imposes recurrent stress on urban transport systems and varies substantially across the day. This study conceptualises urban traffic congestion resilience (UTCR) as an hourly proxy state under recurrent congestion conditions and develops an integrated road-segment–hour framework combining Multi-Criteria Decision Analysis (MCDA) and a Back-Propagation Neural Network (BPNN). MCDA establishes a globally comparable and interpretable proxy baseline, while BPNN provides nonlinear surrogate reconstruction and model-dependent attribution diagnosis; adaptive fusion introduces limited numerical adjustment. Applied to one ordinary weekday in Kunming, China, the results show lower UTCR during morning and evening commuting periods and relatively stable midday conditions. Lower values concentrate in central areas, whereas higher values occur more often in peripheral areas, with significant spatial clustering and corridor-level heterogeneity. Robustness checks confirm the stability of the principal diurnal pattern under alternative fusion intervals and weighting treatments, supporting time-sensitive and corridor-specific traffic management.