A pot trial was set up with nine soil types, each replicated five time
s, in randomized blocks, to investigate the effect of soil properties
on the uptake of Al by soybean plants, The variance of the measurement
s was estimated using five replicates of each soil type, This then ena
bled the statistical significance of the difference between results fo
r each soil type to be established, The analytical variance was determ
ined by two analyses per pot, Systematic errors during the analytical
procedure were estimated by the analysis of reagent blanks and CRMs. A
nalysis of variance (ANOVA) was used to estimate the proportions of to
tal variance contributed by analysis, sampling and geochemistry, Robus
t statistics were required because of the high sensitivity of classica
l ANOVA to a small number of outlying values, due to either technical
or geochemical causes, In the case study described here, traditional a
nalytical quality control apparently gave a satisfactory result, Howev
er, the results from robust ANOVA showed. that the sampling error was
the limiting factor for this application, It is suggested that the env
ironmental analyst must consider errors from procedures of both sampli
ng and analysis of environmental samples.