Structural equation modelling of mercury intra-skeletal variability on archaeological human remains.

Archaeological burial environments are useful archives to investigate the long-term trends and the behaviour of mercury. In order to understand the relationship between mercury, skeletons and soil, we applied Partial Least Squares - Structural Equation Modelling (PLS-SEM) to a detailed, multisampling (n = 73 bone samples +37 soil samples) design of two archaeological graves dating to the 6th to 7th centuries CE (A Lanzada site, NW Spain). Mercury content was assessed using a DMA-80, and data about bone structure and the grave soil/sediments were obtained using FTIR-ATR spectroscopy. The theoretical model is supported by proxies of bone structure, grave soil/sediments, and location of the bone within the skeleton. The general model explained 61 % of mercury variance. Additionally, Partial Least Square - Prediction Oriented Segmentation (PLS-POS) was also used to check for segmentation in the dataset. POS revealed two group of samples depending on the bone phase (hydroxyapatite or collagen) controlling the Hg content, and the corresponding models explained 86 % and 76 % of Hg variance, respectively. The results suggest that mercury behaviour in the graves is complex, and that mercury concentrations were influenced by i) the ante-mortem status of the bone matrix, related to the weight of each bone phase; ii) post-mortem evolution of bone crystallinity, where bone loses mercury with increasing alteration; and iii) the proximity of the skeletal pieces to mercury target organs, as decomposition and collapse of the thoracic and abdominal soft tissues causes a secondary mercury enrichment in bones from the body trunk during early post-mortem. Skeletons provide a source of mercury to the soil whereas soil/sediments contribute little to skeletal mercury content.