Data from: Using artificial intelligence classification of videos to examine the environmental, evolutionary and physiological constraints on provisioning behavior

The use of artificial intelligence (AI) technologies can revolutionize how
we approach data collection and analysis in behavioral ecology. One such
example is in provisioning behavior. Parents of altricial species are
selected to provide parental care (such as food provisioning) for their
offspring, but there is substantial variation in the level of this care.
Provisioning rate may be determined environmentally, by the physiological
ability of parents and needs of nestlings, or by evolutionary
incentives.  We quantified provisioning rate in 20 purple martin
(Progne subis) nests in the context of an experimental reduction of nest
ectoparasites. 10 nests had a parasite reduction treatment, and 10 nests
were controls. By using AI to automate the analysis of nest camera videos
we were able to obtain nearly-continuous provisioning rate information at
a high temporal resolution for the first half of the nestling period. We
used random forest modeling to assess the factors determining provisioning
rate and found evidence for environmental, evolutionary and physiological
constraints and incentives on provisioning. Birds appeared to be
environmentally limited in their provisioning in cool, wet conditions,
especially later in the breeding season; but adjusted their provisioning
according to the changing physiological needs of nestlings. We found
evidence for a compensatory response to increased parasite load, in which
parents increased provisioning to more heavily parasitized nests.