RESUMO

The application of statistical modeling is still infrequent in mercury research in peat, despite the ongoing debate on the weight of the diverse factors (climate, peat decomposition, vegetation changes, etc.) that may affect mercury accumulation. One of the few exceptions is the Hg record of Pinheiro mire (souheast Brazil). Previous studies on this mire modeled mercury using principal components regression and partial least squares. These methods assume independence between factors, which is seldom the case in natural systems, thus hampering the identification of mediating effects and interactions. To overcome these limitations, in this reserach we use structural equation modeling (PLS-SEM) to model mercury and bromine peat records - bromine has been used in some investigations to normalize mercury accumuation. The mercury model explained 83% of the variance and suggested a complex control: increased peat decomposition, dust deposition and humid climates enhanced mercury accumulation, while increased mineral fluxes resulted in a decrease in mercury accumulation. The bromine model explained 90% of the variation in concentrations: increased dust deposition and peat decomposition promoted bromine accumulation, while time (i.e. peat age) promoted bromine depletion. Thus, although mercury and bromine are both organically bound elements with relevant atmospheric cycles the weights of the factors involved in their accumulation differed significantly. Our results suggest caution when using bromine to normalize mercury accumulation. PLS-SEM results indicate a large time dependence of peat decomposition, catchment mineral fluxes, long-term climate change, and atmospheric deposition; while atmospheric dust, mineral fluxes and peat decomposition showed high to moderate climate dependency. In particular, they also point to a relevant role of autogenic processes (i.e. the build up and expansion of the mire within the catchment), which controlled local mineral fluxes; an aspect that has seldom been considered.