Because these processes can all influence wildlife condition differently, the authors make the understandable point that identifying distinct pathways is quite the challenge. Yet by measuring distinct aspects of condition (e.g., mass, fat, muscle scores, cortisol), researchers can begin to disentangle if differences in condition between low- and high-resource habitat are due to energetic constraints or adaptive mass regulation. In other words, if urban wildlife have lower mass and fat but high muscle scores, we might say with some confidence that we're observing the effects of adaptive mass regulation. But if our urban populations show lower fat and muscle, then we'd suggest these animals are energetically and nutritionally limited.
So then the authors applied this multi-dimensional approach to measuring condition among four populations of house sparrows in France that lie along an urbanization gradient. A suite of condition measures was recorded within each population for both juveniles (n = 42) and adult birds (n = 68) to assess if benefits/costs of urbanization affects life history stages differently. Urbanization here was determined by a principal components analysis of five habitat characteristics such as road density and vegetation density. I like the fact that at least some component of basal resources was added here (vegetation), although it would have been nice if some measure related to urban food availability (the supposed driver of the study) was included in the PCA (i.e., density of bird feeders or parks perhaps?).
The statistics here were fairly straightforward. The authors fit GLMs predicting the various condition and physiology measurements as a function of urbanization score (i.e., the four sites), bird age, bird sex, and relevant interaction terms, and then used stepwise model selection to get their lowest AIC model. I understand using a categorical "urbanization score" here given that the study only had four sizes, but if the authors had managed to have a few replicated per urban treatment I wonder if we would have seen any interesting non-linear relationships with a continuous urban score variable. From these data with one replicate each, it's hard to know if any effects are site-specific or actually do the urbanization. Oh well.
Accordingly the authors reject the possibility of hypotheses 2 and 3, in that high food availability leads to poorer condition through adaptive mass regulation or high competition. They also didn't find much benefit in terms of condition from urbanization, which rejects hypothesis 1 (high food = high condition). What's left? The interpretation here that I like returns to the urban birds being smaller in size and mass. Because the tarsus is almost fully grown at fledgling in house sparrows, the smaller size of urban birds could be due to energetic constraints on young, developing birds. The authors offer a few interpretations here, but given that they've ruled out three of the four hypotheses, mechanism four seems to make sense. Namely, urban habitats could provide young birds with poor-quality food (think high carbs, low protein) that impairs early development. This especially could make sense given that young, urban birds had more fat than adult counterparts, which may imply that these birds are being provisioned with non-natural food sources.
I thought this was a nice approach to piecewise challenging varying hypotheses about urban resources and wildlife condition, and the ending point about impaired development offers some interesting thoughts on the long-run fitness consequences of anthropogenic resources. This has been an emerging and conflicting area of research recently, with some high-profile results supporting supplemental feeding (e.g., Vincenzi et al. 2015) while others demonstrating deleterious population outcomes (e.g., Plummer et al. 2013). Although I don't think this study necessarily resolves any of these issues, especially given some of the limitations about sample size and site replication, they do offer a nice multidimensional approach to considering varying aspects of condition and physiology that can highlight different resource/fitness mechanisms. The Meillère study makes me want to go back to my vampire bat data and test if I see differences between the mass ~ livestock density relationship and the forearm length ~ livestock density relationship, rather than looking just at the joint body condition measure of mass~forearm residuals (but you'll have to go to SICB 2016 for that).