This work explores the problem of automatic discovery of different acoustic regions in the world, as experienced by a mobile robot. We use a temporally smoothed variant of Latent Dirichlet Allocation (LDA) to compute both the region models, and most likely region labels associated with each time step in the robot's trajectory. We have experimented with two datasets containing sound recorded from 51 and 43 minute long trajectories through downtown Montreal and the McGill University campus. Our preliminary experiments indicate that the regions discovered by the proposed technique correlate well with ground truth, as labeled by a human expert.