During the last decade there is an upsurge in interest to quantify the relative contributions of ecological and evolutionary processes to observed trait change. Partitioning these contributions is a cornerstone of retrieving better insights in eco-evolutionary dynamics. We compare existing metrics to newly developed metrics to partition ecological and evolutionary drivers of average population trait change that are based on the Price equation and on norms of reaction, and extend them for the analysis of community ecology and of trait variation in space. We highlight advantages and limitations of the different partitioning metrics. For data sets both amenable to the reaction norm and Price partitioning metrics, we develop a new, synthetic metric that combines the strengths of both approaches. Applied at the community level, this metric can partition trait change in components due to phenotypic plasticity within genetic lineages, genetic changes within lineages, genetic changes due to shifts in frequencies of genetic lineages, evolution of plasticity, and components due to frequency shifts of species and to the gain and loss of species in the community. In addition we give empirical examples applicable for the metrics discussed to illustrate the mathematically found differences.
Results/Conclusions
In general we found that the various metrics yielded different estimates of the contribution of ecological and evolutionary processes to trait changes. Most of these differences are due to problems that could already be indicated at the population level. We found that the Price equation yield problematic estimates whenever there is an evolutionary component that cannot in a straightforward way be attributed to lineage sorting. In some cases the problem is limited to an uncertainty on whether the component is due to evolution or to phenotypic plasticity. However, when these processes being grouped in the same component in the Price equation counteract one another, the resulting estimates are wrong. The second metric discussed, the reaction norm approach, calculates the contributions using species averages, and can therefore not quantify lineage sorting in the observed trait change. We also found when comparing the Price equation with the combined Price-Reaction-Norm equation that the species sorting component of the Price equation contains species sorting times evolutionary interactions. When modifying the metrics to analyse spatial data, we found that different options are possible dependent on the research question. As a conclusion we want to stress that researchers need to carefully consider their choice of partitioning metrics and take their limitations into account.