Dolphin populations are often considered an indicator of ocean health, yet they have historically been difficult to monitor. These animals live in remote, variable environments, and spend much of their time out of sight, below the sea surface. Passive acoustic monitoring (PAM) has become a viable method for gathering data on marine mammals. Instruments can be placed on the seafloor for long periods, recording animal sounds in the environment, regardless of oceanographic conditions, time of day and site accessibility. The lingering challenge is in translating acoustic detections into quantitative population density estimates. Density estimation techniques using PAM have been developed for and applied to marine mammals, but they have rarely been used for long term studies using single sensors, or applied to dolphins. The Deepwater Horizon (DWH) oil spill event and its unknown impacts on offshore marine mammals provided an impetus for collecting a long term PAM dataset aimed at monitoring offshore marine mammals in the Gulf of Mexico (GOM). In this work I develop a framework for long-term, high resolution, quantitative monitoring of dolphin populations using this dataset, which stretches over three years at five sites. Delphinid density estimation involves a series of steps: First, the probability of detecting delphinid cues using PAM is estimated, accounting for variability by site, species and season. Cues are then detected in the acoustic recordings, using methods consistent with the constraints of the detection probability estimation process, and classified to species. Finally, density estimates are produced by bringing together the detection counts, probabilities, and species- specific behavioral parameters. We find distinct annual cycles in animal density in the northern GOM with peaks for most species in spring and summer months. Long-term increases in local densities were seen for Stenellid dolphins and pilot whales at sites east of the DWH site, which are not seen at sites to the south and east. This work represents significant progress toward the goal of monitoring dolphin populations with minimal impact, high temporal resolution, and improved accuracy. These methods are broadly applicable to PAM efforts. As peripheral data are improved and expanded, estimates can be refined using this framework