Estimation of plant sampling uncertainty: an example based on chemical analysis of moss samples

In order to estimate the level of uncertainty arising from sampling, 54 samples (primary and duplicate) of the moss species Pleurozium schreberi (Brid.) Mitt. were collected within three forested areas (Wierna Rzeka, Piaski, Pos?owice Range) in the Holy Cross Mountains (south-central Poland). During the fieldwork, each primary sample composed of 8 to 10 increments (subsamples) was taken over an area of 10 m² whereas duplicate samples were collected in the same way at a distance of 1–2 m. Subsequently, all samples were triple rinsed with deionized water, dried, milled, and digested (8 mL HNO₃ (1:1)?+?1 mL 30 % H₂O₂) in a closed microwave system Multiwave 3000. The prepared solutions were analyzed twice for Cu, Fe, Mn, and Zn using FAAS and GFAAS techniques. All datasets were checked for normality and for normally distributed elements (Cu from Piaski, Zn from Pos?owice, Fe, Zn from Wierna Rzeka). The sampling uncertainty was computed with (i) classical ANOVA, (ii) classical RANOVA, (iii) modified RANOVA, and (iv) range statistics. For the remaining elements, the sampling uncertainty was calculated with traditional and/or modified RANOVA (if the amount of outliers did not exceed 10 %) or classical ANOVA after Box-Cox transformation (if the amount of outliers exceeded 10 %). The highest concentrations of all elements were found in moss samples from Piaski, whereas the sampling uncertainty calculated with different statistical methods ranged from 4.1 to 22 %.