Towards an unbiased method for quantifying treatment effects on soil carbon in long-term experiments considering initial within-field variation

Agricultural management effects on soil organic carbon (SOC) can be significant, but are often relatively small compared with the total SOC pool and its field–scale variability. In field experiments, completely uniform initial SOC concentrations across all plots are unlikely. Therefore, measured treatment effects through time are biased by these initial differences. This study questioned how much this bias can represent and whether it can be neglected or must be accounted for. In 1116 investigated pairs of treatments in 10 meta-replicated Swedish long-term soil fertility experiments, the average absolute initial difference in SOC was 1.3 g C kg-1, whereas the average absolute difference after about 50 years of contrasting management was 1.5 g C kg-1. Initial differences in SOC were significantly correlated with final differences (R2 =0.14) and in 47% of all investigated pairs the absolute initial difference was higher than the absolute final difference. However, simple subtraction of initial differences between pairs from final differences to isolate the treatment effect neglect the fact that the soil with the higher initial SOC content will lose proportionally more C through mineralization during the experimental period higher SOC content caused higher SOC mineralisation during the experimental period. Therefore, a first-order kinetic model was used to predict the decrease in initial SOC differences over time as influenced by climate. We derived a generic empirical equation that can be used to account for this severe but inherent problem with plot experiments. According to the equation, 65 and 69% of the initial differences in SOC between treatment pairs persisted after 50 years under the climate conditions in southern and central Sweden, respectively. We conclude that initial differences can be severe and have to be accounted for. The proposed empirical equation can reduce the bias in treatment comparisons for different climatic conditions and contribute to improve the quality of data interpretation from long-term field experiments.