Jonas Ardö

Reliable and effective verification of soil carbon sequestration is required for quantification of project-based greenhouse gas mitigation. Direct soil sampling is necessary for measurements at field level. In this study, soil samples from a semiarid agroecosystem of the Sudan were statistically analyzed to evaluate if changes in soil organic carbon (SOC) over time or space were detectable or not, given a certain sample size. One hundred samples were taken from each of three fields. The data collected did not respect normality, and parametric methods, such as the minimum detectable difference (MDD), could not be used to relate, with confidence, the number of samples required to detect specific changes. For this reason, a method incorporating a nonparametric test into a bootstrap routine was developed to calculate the probabilities that a test will detect the differences between treatment groups for specified sample sizes. The nonparametric approach used is a simple way of relating sample sizes, detectable differences, and the probabilities of detection, making it flexible for any study affected by data nonnormality. Data from different simulated composite sampling schemes were also combined with a bootstrap routine in order to quantify the averages and 95% confidence intervals of the variability estimates (variance), necessary for parametric sample size calculations. The variance decreased almost by half each time the number of cores per samples doubled.