Optimizing sampling allocation for detecting management effects on soil CO2 emissions

Abstract

Measurement of soil CO2 flux is an important tool for detecting managementinduced changes in soil C. The objective of this study was to analyze sources of variability of a recently published CO2 flux dataset to identify a sampling protocol with optimal allocation of replication, sub-sample and treatment numbers for detecting treatment differences. The dataset comprised daily CO2 flux measurements from a long-term study with treatments of conventional tillage (CT) and no-till (NT) under continuous wheat (CONT) and fallow-wheat rotation (F-W) in a randomized complete block design (RCBD) with four blocks. PROC MIXED in SAS was used to estimate variances. The standard error of the difference (SED) between two treatment means was used as the precision indicator. Although increasing the number of replications effectively reduced SED, sub-sampling also often improved detection of treatment differences because sub-sample variance (σ2δ) was higher than experimental unit variance (σ2ε) on most sampling days. When treatments with small CT vs. NT difference were excluded, degrees of freedom for treatment effects were reduced and both variances were generally increased or unchanged, resulting in increased requirement for sub-sampling. Based on the selected dataset, we produced graphs showing the number of days on which a CT vs. NT difference of 0.3 μmol CO2 m-2 s-1 could be detected at P < 0.10 as a function of replication, sub-sample and treatment numbers. This approach may be used as a guide to optimize sample allocation in similar studies, though site- and experiment-specific factors (e.g., spatial and temporal variability of CO 2 flux, size of treatment difference to be detected, the required confidence level) should also be considered.