Sampling optimization at the sampling-site scale for monitoring atmospheric deposition using moss chemistry

Fifty samples of Scleropodium purum were collected from a single study area for individual analysis with the aim of optimizing the protocol for sampling terrestrial mosses in order to biomonitor atmospheric deposition. The concentrations of Ca, Cu, Fe, Hg, Ni, and Zn were determined in each sample and mass-weighted means were calculated. Analyses were also carried out on a composite sample. The distributions of the concentrations and the existence of spatial patterns were studied using semivariograms. The data for elements considered to be contaminants (Cu, Ni and Zn) were log-normally distributed and highly variable. Of the elements studied, only Ni and Zn were spatially dependent, due to atmospheric deposition. The three main sources of error introduced when making composite sample were determined, and possible means of reducing them were suggested. To summarize: (i) measurements were made on a composite sample—not on subsamples—leading to an error that was occasionally >20% (including analytical error) and that was basically unavoidable; (ii) differences in weight between the subsamples led to an error that was scarcely >2% and that could be easily reduced by using subsamples of similar weights; and (iii) sources of variation related to the number of subsamples in a composite sample may lead to an error >30% when the usual methodology is followed. We propose to collect a minimum of 30 subsamples wherever contamination is suspected.