Given well‐known environmental conditions, matched field processing has been shown to be a promising signal processing technique for the localization of acoustic sources. However, when environmental data are incomplete or inaccurate, a ‘‘mismatch’’ occurs between the measured and model fields that can lead to a severe degradation of the localization estimator. The possible mismatch effects of surface and internal waves on matched field processing in a shallow water waveguide were investigated. A modified ray theory was utilized, based on the work of Tindle [J. Acoust. Soc. Am.70, 813–819 (1981);73, 1581–1586 (1983)], to calculate the acoustic pressure field. Thus it was possible to simply incorporate range‐dependent environmental conditions as well as to generalize this work to deeper waveguides. In general, the conventional (Bartlett) matched field beamformer does not provide sufficient resolution to unambiguously locate a source, even in a perfectly matched environment. The maximum likelihood method matched field beamformer has much better resolution, but is extremely susceptible to mismatch. The mismatch due to surface roughness can result in a large reduction of the estimator peak. Part, but not all, of the peak can be regained by (1) using a model that includes incomplete reflection at the surface based on actual sea surface statistics and (2) short‐time averaging of the measured signal, with times on the order of the period of the surface waves. Mismatch due to internal waves can also result in a large degradation of the estimator. Averaging over the same time period as surface waves provides little improvement and leads one to surmise that internal waves may be a limiting constraint on matched field processing. Finally, the surface and internal wave fields were combined with a slowly moving source. This example highlights the necessity for the development of a beamformer that has a broader mainlobe while maintaining adequate sidelobe suppression. This issue is addressed by looking at two such beamformers.