What's novel about the PNAS study here is that the authors allocated experimental food subsidy treatments across 23 residential properties in the city of Aukland. Experimental feeding of birds has been done before in the context of longitudinal, fitness-oriented studies, but to my memory I can't recall a paper that's used a supplemental feeding experiment at this large a scale to infer effects on community structure. Food treatment here consisted of supplying birds at the 11 experimental plots with a daily hearty ration of bread and seed (i.e., abundant and reliable!) and then performing point counts to estimate abundance of each species per month over the annual period and until four months after the experiment (i.e., before, during, and after treatment groups). So a relatively simple (but by no means easy) study design. The analyses followed two paths: multivariate approaches and mixed-effects models. I won't belabor the stats too much, but the authors used a combination of NMDS and PERMANOVA/PERMDISP to test if supplemental feeding and feeding period affected avian community structure. They then tested the effect of feeding treatments on the abundance of each species in turn using GLMM negative binomial models with property (i.e., each site) as a random effect to account for repeated measures. The GLMMs also accounted for spatial autocorrelation in point counts and some covariates such as background feeding frequency and vegetation cover.
An interesting point the authors make is that the increases to house sparrow and spotted dove relative abundance occurred after the breeding season, which points toward other mechanisms of population growth influenced by supplemental feeding, such as boosted immigration. This welcomes some speculation on source–sink dynamics and the role that unmanaged food resources such as bird feeders might have on metapopulation dynamics and the spatial patterns of infectious disease. This has been on my mind for some time now, especially coming off ESA. Amplified immigration into these source patches of high resource availability could increase connectivity and facilitate the movement of infected individuals, which could allow pathogens to spread across urban landscapes (e.g., Hess 1996, Ecology). We're currently working on a metapopulation model that shows the same dynamic can occur through improvements to the persistence time of populations in high-quality patches. Yet since the community composition (and relative abundances of individual species) seemed to equilibrate to control levels after the cessation of feeding, it'll certainly be worth modifying our framework to include between-patch movement heterogeneity. This is an aspect of metapopulation dynamics that I'm thrilled to start tackling and it'll be interesting to see how the dynamics of infection spread change as we build this aspect of resource dependence into our models.
*I also really liked the time series of species richness and bird density as a function of feeding treatment and native/invasive status. These graphs especially demonstrate that supplemental feeding produces a boom in invasive bird species abundance, which then declines almost immediately after the experiment stops.