Designing mixtures of varieties for multifunctional agriculture with the help of ecology. A review

Abstract

The study of natural ecosystems and experiments using mixtures of plant species demonstrates that both species and genetic diversity generally promote ecosystem functioning. Therefore, mixing crop varieties is a promising alternative practice to transform modern high-input agriculture that is associated with a drastic reduction of within-field crop genetic diversity and is widely recognized as unsustainable. Here, we review the effects of mixtures of varieties on ecosystem functioning, and their underlying ecological mechanisms, as studied in ecology and agronomy, and outline how this knowledge can help designing more efficient mixtures. We recommend the development of two complementary strategies to optimize variety mixtures by fostering the ecological mechanisms leading to a positive relationship between biodiversity and ecosystem functioning and its stability through time, i.e., sampling and complementarity effects. (1) In the “trait-blind” approach, the design of high-performance mixtures is based on estimations of the mixing abilities of varieties. While this approach is operational because it does not require detailed trait knowledge, it relies on heavy experimental designs to evaluate mixing ability. (2) The trait-based approach is particularly efficient to design mixtures of varieties to provide particular baskets of services but requires building databases of traits for crop varieties and documenting the relations between traits and services. The performance of mixtures requires eventually to be evaluated in real economic, social, and agronomic contexts. We conclude that the need of a multifunctional low-input agriculture strongly increases the attractiveness of mixtures but that new breeding approaches are required to create varieties with higher mixing abilities, to foster complementarity and selection effects through an increase in the variance of relevant traits and to explore new combinations of trait values.