Integrative exposure assessment through classification and regression trees on bioaccumulation of metals, related sampling site characteristics and ecoregions

The UNECE Heavy Metals in Mosses Surveys measure and spatially predict environmental concentration (PEC) of metals in mosses for ecotoxicological risk assessments. Up to now, no statistical sound investigation was dedicated to those boundary conditions which, aside from the atmospheric depositions of metals from the atmosphere down to the land surface, might influence the bioaccumulation of metals. Thus, the article focuses on the integrative analysis of the data on the bioaccumulation of metals in Germany 1990, 1995 and 2000 on the one hand and of data on conceivable boundary conditions on the other hand. To this end Classification and Regression Trees (CART) were used because CART is a very powerful statistical technique that enables to uncover structures in large data sets containing categorical and continuous data without any transformation of scale dignity. Within the framework of the Metals in Mosses Surveys, moss samples were taken at 592 sites in 1990, at 1026 sites in 1995 and at 1028 sites in 2000 in Germany. At each of them mosses were sampled and the concentrations of up to 40 metals were measured. The sampling, the ecological characteristics of the sampling sites, and the chemical analysis were documented in a metadata base. An ecoregionalisation of Germany was calculated with data on natural vegetation, altitude, soil texture and climate by means of CART. The ecoregions were mapped by GIS and intersected with multi-metal bioaccumulation indices calculated from the monitoring data 1990, 1995 and 2000. These indices were calculated by percentile statistics and the concentrations of As, Cd, Cr, Cu, Fe, Hg, Ni, Pb, V and Zn were integrated. To reach an integrative exposure assessment, for each survey a CART-model was computed encompassing the multi-metal bioaccumulation indices, the metadata and the ecoregionalization. The models describe the multivariate correlations of metal bioaccumulation with site-specific and ecoregional characteristics in a comprehensive and holistic manner over time, space and metal species. This is of crucial importance for the next step in ecological risk assessment, i.e. the interpretation of the exposure data with regard to predicted no effect concentrations (PNEC) and the sensitivity of ecosystems.