Low-power chemical sensors deployed on mobile platforms make it possible to monitor pollutant concentrations across large urban areas. However, chemical sensors are prone to drift (e.g., aging, damage, poisoning) and have to be calibrated periodically. In this paper, we present an opportunistic calibration approach that relies on encounters between sensors; when in vicinity of each other, sensors exchange measurements and use the accumulated information to re-calibrate. We formulate the calibration process as weighted least-squares, where the most recent measurements are assigned the highest weights. We model the weights with an exponential decay function (in time) and optimize the decay constant using simulated annealing (SA). We validated the proposed method on a simulated sensor network with the sensors' mobility driven by random-waypoint (RWP) models. We present results in terms of average calibration errors for different weight functions, and network sizes.