Plant monocultures growing for extended periods face severe losses of productivity. This phenomenon, known as ‘yield decline’, is often caused by the accumulation of above- and belowground plant antagonists. The effectiveness of plant defences against antagonists might help explaining differences in yield decline among species. Using a trait-based approach, we studied the role of 20 physical and chemical defence traits of leaves and fine roots on yield decline of 18-year old monocultures of 27 grassland species. We hypothesized that yield decline is lower for species with high defences, that root defences are better predictors of yield decline than leaf defences, and that in roots, physical defences better predict yield decline than chemical defences, while the reverse is true for leaves. We additionally hypothesized that species increasing the expression of defence traits after long-term monoculture growth would suffer less yield decline. We summarized leaf and fine root defence traits using principal component analysis and analysed the relationship between defence traits mean as a measure of defence strenght and defence traits temporal changes of the most informative components and monoculture yield decline. The only significant predictors of yield decline were the mean and temporal changes of the component related to specific root length and root diameter (e.g. the so called collaboration gradient of the root economics space). The principal component analysis of the remaining traits showed strong trade-offs between defences suggesting that different plant species deploy a variety of strategies to defend themselves. This diversity of strategies could preclude the detection of a generalized correlation between the strength and temporal changes of defence gradients and yield decline. Our results show that yield decline is strongly linked to belowground processes particularly to root traits. Further studies are needed to understand the mechanism driving the effect of the collaboration gradient on yield decline.

Ecosystem management aims at providing many ecosystem services simultaneously. Such ecosystem multifunctionality can be limited by trade-offs and increased by synergies among the underlying ecosystem functions (EF), which need to be understood to develop targeted management. Previous studies found differences in the correlation between EFs. We hypothesised that correlations between EFs are variable even under the controlled conditions of a field experiment and that seasonal and annual variation, plant species richness, and plot identity (identity effects of plant communities such as the presence and absence of functional groups and species) are drivers of these correlations. We used data on 31 EFs related to plants, consumers, and physical soil properties that were measured over 5 to 19 years, up to three times per year, in a temperate grassland experiment with 80 different plots, constituting six sown plant species richness levels (1, 2, 4, 8, 16, 60 species). We found that correlations between pairs of EFs were variable, and correlations between two particular EFs could range from weak to strong correlations or from negative to positive correlations among the repeated measurements. To determine the drivers of pairwise EF correlations, the covariance between EFs was partitioned into contributions from plant species richness, plot identity, and time (including years and seasons). We found that most of the covariance for synergies was explained by species richness (26.5%), whereas for trade-offs, most covariance was explained by plot identity (29.5%). Additionally, some EF pairs were more affected by differences among years and seasons and therefore showed a higher temporal variation. Therefore, correlations between two EFs from single measurements are insufficient to draw conclusions on trade-offs and synergies. Consequently, pairs of EFs need to be measured repeatedly under different conditions to describe their relationships with more certainty and be able to derive recommendations for the management of grasslands.