Transport assessment plays a vital role in urban planning and traffic control, which are influenced by multi-faceted traffic factors involving road infrastructure and traffic flow. Conventional solutions can hardly meet the requirements and expectations of domain experts. In this paper we present a data-driven solution by leveraging a visual analysis system to evaluate the real traffic situations based on taxi trajectory data. A sketch-based visual interface is designed to support dynamic query and visual reasoning of traffic situations within multiple coordinated views. In particular, we propose a novel road-based query model for analysts to interactively conduct evaluation tasks. This model is supported by a bi-directional hash structure, TripHash, which enables real-time responses to the data queries over a huge amount of trajectory data. Case studies with a real taxi GPS trajectory dataset (> 30GB) show that our system performs well for on-demand transport assessment and reasoning.