Reduced sampling frequency can still detect changes in abundance and phenology of migratory landbirds

Many monitoring programs, such as bird banding stations, would benefit from improvements in their sampling designs. Autumn bird migration records from Manomet in southeastern Massachusetts from 1969 to 2012, collected 5 days a week, every year, show strong evidence of changes in bird migration cohort sizes and average arrival dates. Using these previously identified trends for five common species, we subsampled days of the week and total years to simulate different sampling frequencies, and we varied the length of the study by subsampling time periods. Even with 60% reduced sampling from five days to just two days per week, trends in abundance were still statistically detectable and most patterns of changing phenology were still present. Sampling every other year had no effect on the ability to detect changing abundance or phenology, and most of the effects were still detected with sampling every third or fourth year. The observed trends in abundance and phenology for the full 44-year data set were not statistically detectable in simulated shorter study durations of 22, 15 and 11 years. These results suggest that sampling over long periods using constant methods is key to detecting changes over time in abundance and phenology; in many cases these methods can be made more efficient by reducing sampling intensity without losing the ability to detect trends. This study also provides corrected results for an earlier study on autumn migration times by Ellwood et al. (2015), which incorporate recently identified changes in sampling design.