The Crop Rotation Effect: Empirical and Simulated Spatiotemporal Yield Variation.

Annual and cumulative mean yield, yield variance, and total yield of four phases of a 2-yr crop rotation and two phases of a 4-yr crop rotation nested within conventional and organic cropping systems were subjected to geostatistical and generalized linear mixed model analyses with repeated measurements to predict, then simulate temporal yield variation under projected climate change. Soil physicochemical attributes were measured and used as covariates where applicable. Spatial variance, accounted for by regular (rows and columns) or irregular (plot coordinates) grid, was estimated in the presence of temporal variance, which was accounted for by the interaction of years (i.e., time in repeated measurements) with fixed factors (cropping systems, crop rotations, rotation phases, crops, and management inputs) using generalized linear mixed models. Six covariance models (uniform, first and second order autoregressive, first and second order antedependence, and unstructured) were tested under restricted maximum likelihood (REML) estimation and the model with the smallest deviance was selected to approximate the relationship between observed data associated with the experimental units. Results of the simulated internal variability due to cyclical pattern of crop rotations and external inputs were used to formulate a decision support system for structuring the temporal succession of crops and determine the optimum long-term proportion of annual and perennial crops in flexible crop rotations under conventional and organic cropping systems in response to projected climate change.