Automated acoustic recorders were deployed at 331 sites across the Mojave Desert ecoregion and 263 sites across the Great Valley ecoregion between March and July of 2016 and 2017. At each survey location, an SM3-BAT bioacoustic recorder was deployed with both an SMM-A1 acoustic microphone and an SMM-U1 ultrasonic microphone (Wildlife Acoustics, Inc., Maynard, MA, USA, hereafter termed ARU). Automated recorders were cable locked to securely-placed T-posts, with microphones 2 meters above the ground, and were located within 10 meters of the site center at a relatively open spot not occupied by thick vegetation. The ARUs were programmed to record three 5-minute sessions each morning during the survey period, with the first session at 30 minutes before sunrise, the second at sunrise, and the third at 30 minutes after sunrise. The ARUs were also programmed to alternate between 7.5-minute audible acoustic recordings and 22.5-minute triggered, full-spectrum ultrasonic recordings from 30 minutes before sunset until 0400 the following morning, at which time the ultrasonic recordings ceased; the 7.5-minute acoustic recordings continued every 30 minutes until the 5-minute dawn chorus recordings began at 30 minutes before sunrise. Deployments lasted for seven consecutive days, on average. For the avian acoustic analysis, all birds were surveyed on the acoustic channel from the three 5-minute dawn chorus recordings collected from three mornings during the survey period, preferably from consecutive days. Recordings were reviewed, and bird species identified by song or call. To aid in bird identification, spectrograms were visually examined using Raven Pro software (v. 1.5; Cornell Lab of Ornithology Bioacoustics Research Program, Ithaca, NY, USA). Multispecies hierarchical occupancy models were then used to estimate 1) the probability a given bird species occurred within the area sampled by an acoustic recorder during the survey period, 2) the richness of bird species, and 3) community and species-specific responses to different ecological covariates. For each sampling location, each 5-minute acoustic recording (n = 9) was treated as a repeat survey at that particular site. Occurrence and detection probabilities were assumed to differ between years and among species, and were assumed to be influenced by ecological covariates (e.g., crop cover, forest cover, water availability).
