Fig. 1: Model-predicted effects of functional richness (FR) on annual crop rotation outputs when forage crops are used for milk production.

a–d, Calories (a) and macronutrients (carbohydrates (b), proteins (c) and fats (d)) produced per year by all the crops in the rotation, when they cover 1 ha. FR is the number of functional types included in the rotation, with FR 1M referring to a cereal monoculture and FR 1C to a cereal-only rotation. The statistical models were obtained by fitting the 12,517 whole-rotation outputs relative to the 16 long-term experiments (see Supplementary Table 1 for details). Horizontal dashed lines are model predictions relative to FR 1M at time 0. Filled symbols refer to model predictions after 5 years (orange squares), 10 years (green circles) and 20 years (blue triangles) following the implementation of the rotation—times arbitrarily chosen to show changes over time. Whiskers extend over the 5–95% confidence intervals. Symbols indicating significances refer to contrasts across levels of FR within each time, using FR 1M at the same time as the baseline (+, P < 0.05; ++, P < 0.01; +++, P < 0.001), and across time within each FR with time 0 at that same FR as the baseline (^, P < 0.05; ^^, P < 0.01; ^^^, P < 0.001). Post hoc two-sided tests were adjusted for multiplicity using multivariate t-distribution. Note the difference in the y axis scale among plots.