Urban Sensing employs physical-world mining to create a digital model of the physical world using a large number of sensors. Handling the large amount of data generated by sensors is costly and requires energy-saving measures for sensing and sensor data transmission. Such schemes often affect data quality and message delay. However, the detection of real-world situations using Complex Event Processing on sensor data has to be dependable and timely and requires precise data. In this position paper, we propose an approach to integrate the contradicting optimization goals of energy-efficient wireless sensor networks and dependable situation detection. It separates the system into the following tiers: First, to support energy-efficiency and allow sparse, unconnected sensor networks, we exploit the mobility of people through Delay Tolerant Networking for collecting sensor data. This frees sensor nodes from energy-expensive routing. Second, we employ Diagnostic Simulation which provides data that is complete, precise and in time and therefore supports quality-aware situation detection.