How Many Replications Are Needed for Comparing Whole Profile Soil C Stocks?.

Soil carbon (C) stock assessments for the whole soil profile are of great importance however their measurements are highly variable. The outcome of the high variability is that often no statistically significant difference between the studied treatments can be detected in a course of statistical analyses. A common mistake is to interpret this lack of statistical significance as a reason to accept the hypothesis of no difference between the treatments. Our objective is to remind the researchers that inadequate replication is one of the major reasons for the lack of statistical significance; and that a comprehensive power analysis is needed in order to be able to conclude what caused the lack of statistical significance - an absence of meaningful differences or an insufficient replication. Two published studies reporting soil C stocks were used as a source of illustration data. Based on the data from these studies it appears that due to very high variability in soil C in deep soil and in the whole profile C stocks, the probabilities of detecting even very large differences between the treatments are quite low unless the number of soil samples is extremely (most likely prohibitively) large. Therefore, the lack of statistical significance in the deep layer or whole profile C data is often a result of inadequate sampling size and should not be interpreted as the absence of the practically significant differences without a comprehensive post-hoc power analysis. Instead, statistical analysis of soil C stocks should be conducted separately for each sampled horizon and the conclusion about the profile as a whole should be made based on the individual horizon results. If the statistical analysis of the whole profile stock is still believed to be necessary, devising an optimal sampling strategy, such that allows minimizing the overall sampling costs, is highly recommended.