Maximizing Nest Box Monitoring Effort to Detect American Kestrel Site Occupancy

Abstract

ABSTRACT When designing nest monitoring protocols for raptors, investigators face a tradeoff between the desire to detect all nesting attempts and limited time and resources. American Kestrel (Falco sparverius) nest box occupancy (i.e., traditionally defined as number of nest boxes used divided by number of nest boxes available) is often the principal metric for tracking kestrel population size and trend. However, traditional nest box monitoring can lead to underestimates of local population size when kestrel pairs establish territories at nest boxes but are undetected because they do not lay eggs or their nests fail early in the nesting cycle. We analyzed empirical data collected during frequent visits (1–7 d intervals) to nest boxes throughout the breeding season (March–June during 2008–2014) in a dynamic occupancy modeling framework to assess the timing and intensity of monitoring needed to detect (with the same confidence level provided by frequent visits) the presence of kestrels on nest box sites in Florida. Modeled estimates of occupancy were similar to observed rates but trended slightly higher, especially in years with infrequent monitoring. Detection probability varied markedly over the duration of the breeding season; therefore, using a constant detection probability to determine the minimum number of visits needed to detect kestrel presence at a 95% confidence level produced misleading results. Modeling results indicated that >3 nest box inspections per season did not improve estimates of nest box occupancy. The common practice of monitoring American Kestrel nest boxes at approximately monthly intervals appears sufficient to detect the percentage of nest box sites (rather than nest boxes) used by kestrels, provided that monitoring visits are timed to straddle the peak period of egg laying. We recommend three nest box visits during the breeding season in Florida coincident with mid-March, mid-April, and mid-to-late May to maximize detection of kestrel nests. Investigators in other areas of the species' range will need to adjust the timing of our recommendations to local phenology. In addition, if accurately tracking local kestrel population size and trend is a project goal, then studies would benefit from using an occupancy approach rather than simply recording the percentage of nest boxes with nests, especially in years when fewer visits are possible.