1. Altered animal distributions are a consequence of human expansion and development. Anthropogenic noise can be an important predictor of abundance declines near human infrastructure, yet more information is needed to understand noise impacts at the spatial and temporal scales necessary to alter populations. 2. Energy development and associated anthropogenic noise are globally pervasive, and expanding. For example, 600,000 new natural gas wells have been drilled across central North America in less than twenty years. 3. We experimentally broadcast energy sector noise (recordings of compressor engines) in Southwest Idaho (USA). We placed arrays of speakers creating a "phantom natural gas field" in a large-scale experiment, and tested the effects of noise alone on breeding songbird abundance. To examine variation in human-caused noise, we broadcast two types of compressor noise, one with a slightly higher sound intensity and greater bandwidth than the other. 4. Our phantom natural gas field encompassed approximately 100 km2. We broadcast noise for over three continuous months, for each of two seasons, and quantified over 20,000 hours of background sound levels. 5. Brewer's sparrows (Spizella breweri) were affected by our narrowband playback, declining 30% 50 m from the speaker arrays. During our broadband playback, all species combined and Brewer's sparrows decreased 20% and 33% respectively at the scale of our sites (~0.5 km2; up to 400 m from speaker arrays). 6. Our results show the importance of incorporating the acoustic structure of noise when estimating the cost of noise exposure for populations and suggest an urgent need for noise mitigation, such as quieting compressor station noise, in energy extraction fields and natural areas broadly.