Mobile Wireless Sensor Networks (WSNs) hold the potential to constitute a real game changer for our understanding of urban air pollution, through a significant augmentation of spatial resolution in measurement. However, temporal drift, crosssensitivity and effects caused by varying environmental conditions (e.g., temperature) in low-cost chemical sensors (typically used in mobile WSNs) pose a tough challenge for reliable calibration. Based on state-of-the-art rendezvous calibration methods, we propose a novel model-based method for automatically estimating the baseline and gain characteristics of low-cost chemical sensors taking temporal drift and temperature dependencies of the sensors into account. The performance of our algorithm is evaluated using data gathered by our long-term mobile sensor network deployment, developed within the Nano-Tera.ch OpenSense II project in Lausanne, Switzerland. We show that, in a realistic context of sparse and irregular rendezvous events, our method consistently improves rendezvous calibration performance for single-hop online calibration.