A method for soil management assessment in an unreplicated commercial field

Abstract

Context. Unreplicated trials are common in agriculture. However, statistical inferences differ from those of traditional experiments based on small, replicated plots. Aims. To present a method to assess management effects on soil carbon (C) storage from unreplicated, side-by-side field trials. Methods. Two estimates of means with spatially correlated errors are compared using a corrected t-statistic. Then causal inference is made by analysing a significant difference between the means (P < 0.05) and its changes over time. The use of the method is described in comparing yield and organic C stocks between two large fields. Yield was measured during 1997-2005 with a commercial yield monitor, and soil organic C stocks during 2003-2005. The fields experienced the same tillage practice until autumn 2003 and then with different tillage intensity. Key results. The results show that crop C yield did not differ between the fields when using the same tillage practice but was greater in the tilled than the no-till field. The results also suggest that total and particulate organic matter-C contents depend on tillage history. For comparative purposes, the data were also analysed using standard mixed model analysis with a semivariogram model for spatial autocorrelation among the residuals. The mixed model results were generally similar to those of the corrected t-statistic method. The mixed model was often, but not always, less conservative than the corrected t-statistic model. Conclusions. The method allows analysis of whole-field data and improves our understanding of soil C processes in commercial fields, where agricultural assessment cannot involve replication due to agronomic and economic constraints.