Assessing temporal trends of trace metal concentrations in mosses over France between 1996 and 2011: A flexible and robust method to account for heterogeneous sampling strategies.

Air quality biomonitoring has been successfully assessed using mosses for decades in Europe, particularly regarding heavy metals (HM). Assessing robust temporal variations of HM concentrations in mosses requires to better understand to what extent they are affected by the sampling protocol and the moss species. This study used the concentrations of 14 elements measured during four surveys over 15 years in France. Analyses of variance (ANOVA) and a modeling approach were used to decipher temporal variations for each element and adjust them with parameters known to affect concentrations. ANOVA followed by post hoc analyses did not allow to estimate clear trends. A generalized additive mixed modeling approach including the sampling period, the collector and the moss species, plus quadratic effects, was used to analyze temporal variations on repeated sampling sites. This approach highlighted the importance of accounting for non-linear temporal variations in HM, and adjusting for confounding factors such as moss species, species-specific differences between sampling periods, collector and methodological differences in sampling campaigns. For instance, lead concentrations in mosses decreased between 1996 and 2011 following quadratic functions, with faster declines for the most contaminated sites in 1996. On the other hand, other HM showed double trends with U-shaped or hill-shaped curves. The effect of the moss was complex to handle and our results advocate for using one moss species by repeated site to better analyze temporal variations.