Modeling yield Sustainability for different rotations in long-term barley trials

In crop rotation trials, the errors arising over time on the same plot are correlated. Often-used analyses assume zero or constant correlation between the errors while the actual covariance structure for the plot errors might be very different. The objective of this study was to identify the most suitable covariance structure and incorporate the error structure in assessment of the crop rotations in terms of yield productivity and yield sustainability. A set of five covariance structures were examined for barley yield data from a 14-year, two-course barley rotation trial conducted at two locations in northern Syria. Selection of the covariance structures was based on the Akaike information criterion (AIC, Akaike 1974) (a function of penalized log-likelihood) obtained from fitting the structure. Covariance structure with heterogeneous variances and with constant correlation between errors over cycles within the same plot was found to account for most variability in grain and straw yields at both locations. Modeling data with this covariance structure, the legume rotations gave higher productivity as well as higher annual increases compared with the continuous barley system. This implies that an agricultural production system based on a legume following a cereal is likely to be more sustainable for cereal production compared to cereal followed by cereals.