The incorporation of propagation modeling into array signal processing, usually referred to as matched-field processing, has been shown to provide accurate source localization information in many situations. It is well known though that the performance of these methods is quite sensitive to uncertainty about knowledge of the environmental parameters. In this paper, broadband matched-field processing methods that are tolerant of the uncertainty about environmental knowledge are investigated. A ‘‘robust’’ matched-field matched-filter processor based on minimax robust filtering methods is developed. The performance of the robust processor, in the presence of uncertain environmental knowledge, is compared with conventional methods. In the context of a shallow-water example, with uncertain knowledge about the sound-speed profile, the minimax robust processor is shown to provide a considerable improvement in performance with respect to conventional nonrobust methods (i.e., reduced error in source location estimate and increased source peak-to-sidelobe ratio).