Spatial scale and site fidelity shape the correlation between blood and feather mercury in a riverine bird

Blood and feathers are the non-lethal matrices most used to monitor mercury (Hg) in birds. Blood reflects recent intake at the sampling site, whereas feathers record exposure during moult, possibly elsewhere. Their correspondence should therefore vary with spatial scale and with site-fidelity. We tested this using 229 paired red blood cell and feather samples from 199 adult white-throated dippers (Cinclus cinclus) of mixed sex and age on eight rivers in the Chartreuse Mountains, France. Bivariate Bayesian mixed models estimated blood–feather correlations in Hg concentrations at the population level, among and within rivers, and by site fidelity. River explained about half the variance in both matrices, whereas territory explained more variance in blood than in feather Hg (intra-class correlation 0.24 vs. 0.03). Blood and feather Hg were moderately correlated (rpop = 0.45 [95% highest density interval 0.31, 0.59]), combining strong correspondence among rivers (rriver = 0.88 [0.60, 1.00]) with weak correspondence within rivers (rwithin-river = 0.09 [−0.07, 0.27]). Correspondence was stronger in site-faithful than in other birds (rpop = 0.67 vs. 0.19). Within rivers, breeding partners were strongly correlated in blood but not in feather Hg (rwithin-river = 0.70 vs. 0.14). Concentrations in blood Hg, but not feather Hg, declined by 56% within four months. Feathers suit comparisons among rivers or regions, particularly in site-faithful adults, whereas blood is better suited to territories and to birds of unknown moult location, provided that sampling date is standardised or modelled. Blood–feather correlations should be reported at the spatial scale of inference.