A collaborative algorithm is developed to track the channel gains from arbitrary positions in a geographical area to each node of a cognitive radio network. Spatio-temporal shadow fading effects are characterized using an experimentally verified spatial loss field model. Kriged Kalman filtering (KKF) is then applied to track the time-varying shadowing field. The proposed KKF algorithm consists of a distributed Kalman filter that estimates the spatio-temporal trend field, and a Kriging interpolator which captures the temporally white yet spatially descriptive component at the individual sensors. Numerical tests demonstrate that the collaborative tracking algorithm outperforms a non-collaborative alternative in terms of mean-square error when applied to a cognitive radio sensing task.